* [PATCH v2 01/36] include/qemu/atomic128: Support 16-byte atomic read/write for Intel AVX
2022-10-21 7:15 [PATCH v2 00/36] tcg: Support for Int128 with helpers Richard Henderson
@ 2022-10-21 7:15 ` Richard Henderson
2022-10-21 7:15 ` [PATCH v2 02/36] tcg: Tidy tcg_reg_alloc_op Richard Henderson
` (34 subsequent siblings)
35 siblings, 0 replies; 43+ messages in thread
From: Richard Henderson @ 2022-10-21 7:15 UTC (permalink / raw)
To: qemu-devel
Intel has now given guarantees about the atomicity of SSE read
and write instructions on cpus supporting AVX. We can use these
instead of the much slower cmpxchg16b.
Derived from https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104688
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
include/qemu/atomic128.h | 44 ++++++++++++++++++++++++++
util/atomic128.c | 67 ++++++++++++++++++++++++++++++++++++++++
util/meson.build | 1 +
3 files changed, 112 insertions(+)
create mode 100644 util/atomic128.c
diff --git a/include/qemu/atomic128.h b/include/qemu/atomic128.h
index adb9a1a260..d179c05ede 100644
--- a/include/qemu/atomic128.h
+++ b/include/qemu/atomic128.h
@@ -127,6 +127,50 @@ static inline void atomic16_set(Int128 *ptr, Int128 val)
: [l] "r"(l), [h] "r"(h));
}
+# define HAVE_ATOMIC128 1
+#elif !defined(CONFIG_USER_ONLY) && defined(__x86_64__)
+/*
+ * The latest Intel SDM has added:
+ * Processors that enumerate support for Intel® AVX (by setting
+ * the feature flag CPUID.01H:ECX.AVX[bit 28]) guarantee that the
+ * 16-byte memory operations performed by the following instructions
+ * will always be carried out atomically:
+ * - MOVAPD, MOVAPS, and MOVDQA.
+ * - VMOVAPD, VMOVAPS, and VMOVDQA when encoded with VEX.128.
+ * - VMOVAPD, VMOVAPS, VMOVDQA32, and VMOVDQA64 when encoded
+ * with EVEX.128 and k0 (masking disabled).
+ * Note that these instructions require the linear addresses of their
+ * memory operands to be 16-byte aligned.
+ *
+ * We do not yet have a similar guarantee from AMD, so we detect this
+ * at runtime rather than assuming the fact when __AVX__ is defined.
+ */
+extern bool have_atomic128;
+
+static inline Int128 atomic16_read(Int128 *ptr)
+{
+ Int128 ret;
+ if (have_atomic128) {
+ asm("vmovdqa %1, %0" : "=x" (ret) : "m" (*ptr));
+ } else {
+ ret = atomic16_cmpxchg(ptr, 0, 0);
+ }
+ return ret;
+}
+
+static inline void atomic16_set(Int128 *ptr, Int128 val)
+{
+ if (have_atomic128) {
+ asm("vmovdqa %1, %0" : "=m" (*ptr) : "x" (val));
+ } else {
+ Int128 old = *ptr, cmp;
+ do {
+ cmp = old;
+ old = atomic16_cmpxchg(ptr, cmp, val);
+ } while (old != cmp);
+ }
+}
+
# define HAVE_ATOMIC128 1
#elif !defined(CONFIG_USER_ONLY) && HAVE_CMPXCHG128
static inline Int128 atomic16_read(Int128 *ptr)
diff --git a/util/atomic128.c b/util/atomic128.c
new file mode 100644
index 0000000000..55863ce9bd
--- /dev/null
+++ b/util/atomic128.c
@@ -0,0 +1,67 @@
+/*
+ * Copyright (C) 2022, Linaro Ltd.
+ *
+ * License: GNU GPL, version 2 or later.
+ * See the COPYING file in the top-level directory.
+ */
+#include "qemu/osdep.h"
+#include "qemu/atomic128.h"
+
+#ifdef __x86_64__
+#include "qemu/cpuid.h"
+
+#ifndef signature_INTEL_ecx
+/* "Genu ineI ntel" */
+#define signature_INTEL_ebx 0x756e6547
+#define signature_INTEL_edx 0x49656e69
+#define signature_INTEL_ecx 0x6c65746e
+#endif
+
+/*
+ * The latest Intel SDM has added:
+ * Processors that enumerate support for Intel® AVX (by setting
+ * the feature flag CPUID.01H:ECX.AVX[bit 28]) guarantee that the
+ * 16-byte memory operations performed by the following instructions
+ * will always be carried out atomically:
+ * - MOVAPD, MOVAPS, and MOVDQA.
+ * - VMOVAPD, VMOVAPS, and VMOVDQA when encoded with VEX.128.
+ * - VMOVAPD, VMOVAPS, VMOVDQA32, and VMOVDQA64 when encoded
+ * with EVEX.128 and k0 (masking disabled).
+ * Note that these instructions require the linear addresses of their
+ * memory operands to be 16-byte aligned.
+ *
+ * We do not yet have a similar guarantee from AMD, so we detect this
+ * at runtime rather than assuming the fact when __AVX__ is defined.
+ */
+bool have_atomic128;
+
+static void __attribute__((constructor))
+init_have_atomic128(void)
+{
+ unsigned int a, b, c, d, xcrl, xcrh;
+
+ __cpuid(0, a, b, c, d);
+ if (a < 1) {
+ return; /* AVX leaf not present */
+ }
+ if (c != signature_INTEL_ecx) {
+ return; /* Not an Intel product */
+ }
+
+ __cpuid(1, a, b, c, d);
+ if ((c & (bit_AVX | bit_OSXSAVE)) != (bit_AVX | bit_OSXSAVE)) {
+ return; /* AVX not present or XSAVE not enabled by OS */
+ }
+
+ /*
+ * The xgetbv instruction is not available to older versions of
+ * the assembler, so we encode the instruction manually.
+ */
+ asm(".byte 0x0f, 0x01, 0xd0" : "=a" (xcrl), "=d" (xcrh) : "c" (0));
+ if ((xcrl & 6) != 6) {
+ return; /* AVX not enabled by OS */
+ }
+
+ have_atomic128 = true;
+}
+#endif /* __x86_64__ */
diff --git a/util/meson.build b/util/meson.build
index 5e282130df..4b29b719a8 100644
--- a/util/meson.build
+++ b/util/meson.build
@@ -2,6 +2,7 @@ util_ss.add(files('osdep.c', 'cutils.c', 'unicode.c', 'qemu-timer-common.c'))
if not config_host_data.get('CONFIG_ATOMIC64')
util_ss.add(files('atomic64.c'))
endif
+util_ss.add(when: 'CONFIG_SOFTMMU', if_true: files('atomic128.c'))
util_ss.add(when: 'CONFIG_POSIX', if_true: files('aio-posix.c'))
util_ss.add(when: 'CONFIG_POSIX', if_true: files('fdmon-poll.c'))
if config_host_data.get('CONFIG_EPOLL_CREATE1')
--
2.34.1
^ permalink raw reply related [flat|nested] 43+ messages in thread* [PATCH v2 02/36] tcg: Tidy tcg_reg_alloc_op
2022-10-21 7:15 [PATCH v2 00/36] tcg: Support for Int128 with helpers Richard Henderson
2022-10-21 7:15 ` [PATCH v2 01/36] include/qemu/atomic128: Support 16-byte atomic read/write for Intel AVX Richard Henderson
@ 2022-10-21 7:15 ` Richard Henderson
2022-10-25 15:26 ` Philippe Mathieu-Daudé
2022-10-21 7:15 ` [PATCH v2 03/36] tcg: Introduce paired register allocation Richard Henderson
` (33 subsequent siblings)
35 siblings, 1 reply; 43+ messages in thread
From: Richard Henderson @ 2022-10-21 7:15 UTC (permalink / raw)
To: qemu-devel
Replace goto allocate_in_reg with a boolean.
Remove o_preferred_regs which isn't used, except to copy.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
tcg/tcg.c | 45 +++++++++++++++++++++------------------------
1 file changed, 21 insertions(+), 24 deletions(-)
diff --git a/tcg/tcg.c b/tcg/tcg.c
index c9e664ee31..660d5eb098 100644
--- a/tcg/tcg.c
+++ b/tcg/tcg.c
@@ -3606,7 +3606,8 @@ static void tcg_reg_alloc_op(TCGContext *s, const TCGOp *op)
/* satisfy input constraints */
for (k = 0; k < nb_iargs; k++) {
- TCGRegSet i_preferred_regs, o_preferred_regs;
+ TCGRegSet i_preferred_regs;
+ bool allocate_new_reg;
i = def->args_ct[nb_oargs + k].sort_index;
arg = op->args[i];
@@ -3621,9 +3622,12 @@ static void tcg_reg_alloc_op(TCGContext *s, const TCGOp *op)
continue;
}
- i_preferred_regs = o_preferred_regs = 0;
+ reg = ts->reg;
+ i_preferred_regs = 0;
+ allocate_new_reg = false;
+
if (arg_ct->ialias) {
- o_preferred_regs = op->output_pref[arg_ct->alias_index];
+ i_preferred_regs = op->output_pref[arg_ct->alias_index];
/*
* If the input is readonly, then it cannot also be an
@@ -3632,30 +3636,23 @@ static void tcg_reg_alloc_op(TCGContext *s, const TCGOp *op)
* register and move it.
*/
if (temp_readonly(ts) || !IS_DEAD_ARG(i)) {
- goto allocate_in_reg;
+ allocate_new_reg = true;
+ } else if (ts->val_type == TEMP_VAL_REG) {
+ /*
+ * Check if the current register has already been
+ * allocated for another input.
+ */
+ allocate_new_reg = tcg_regset_test_reg(i_allocated_regs, reg);
}
-
- /*
- * Check if the current register has already been allocated
- * for another input aliased to an output.
- */
- if (ts->val_type == TEMP_VAL_REG) {
- reg = ts->reg;
- for (int k2 = 0; k2 < k; k2++) {
- int i2 = def->args_ct[nb_oargs + k2].sort_index;
- if (def->args_ct[i2].ialias && reg == new_args[i2]) {
- goto allocate_in_reg;
- }
- }
- }
- i_preferred_regs = o_preferred_regs;
}
- temp_load(s, ts, arg_ct->regs, i_allocated_regs, i_preferred_regs);
- reg = ts->reg;
+ if (!allocate_new_reg) {
+ temp_load(s, ts, arg_ct->regs, i_allocated_regs, i_preferred_regs);
+ reg = ts->reg;
+ allocate_new_reg = !tcg_regset_test_reg(arg_ct->regs, reg);
+ }
- if (!tcg_regset_test_reg(arg_ct->regs, reg)) {
- allocate_in_reg:
+ if (allocate_new_reg) {
/*
* Allocate a new register matching the constraint
* and move the temporary register into it.
@@ -3663,7 +3660,7 @@ static void tcg_reg_alloc_op(TCGContext *s, const TCGOp *op)
temp_load(s, ts, tcg_target_available_regs[ts->type],
i_allocated_regs, 0);
reg = tcg_reg_alloc(s, arg_ct->regs, i_allocated_regs,
- o_preferred_regs, ts->indirect_base);
+ i_preferred_regs, ts->indirect_base);
if (!tcg_out_mov(s, ts->type, reg, ts->reg)) {
/*
* Cross register class move not supported. Sync the
--
2.34.1
^ permalink raw reply related [flat|nested] 43+ messages in thread* Re: [PATCH v2 02/36] tcg: Tidy tcg_reg_alloc_op
2022-10-21 7:15 ` [PATCH v2 02/36] tcg: Tidy tcg_reg_alloc_op Richard Henderson
@ 2022-10-25 15:26 ` Philippe Mathieu-Daudé
0 siblings, 0 replies; 43+ messages in thread
From: Philippe Mathieu-Daudé @ 2022-10-25 15:26 UTC (permalink / raw)
To: Richard Henderson, qemu-devel
On 21/10/22 09:15, Richard Henderson wrote:
> Replace goto allocate_in_reg with a boolean.
> Remove o_preferred_regs which isn't used, except to copy.
>
> Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
> ---
> tcg/tcg.c | 45 +++++++++++++++++++++------------------------
> 1 file changed, 21 insertions(+), 24 deletions(-)
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
^ permalink raw reply [flat|nested] 43+ messages in thread
* [PATCH v2 03/36] tcg: Introduce paired register allocation
2022-10-21 7:15 [PATCH v2 00/36] tcg: Support for Int128 with helpers Richard Henderson
2022-10-21 7:15 ` [PATCH v2 01/36] include/qemu/atomic128: Support 16-byte atomic read/write for Intel AVX Richard Henderson
2022-10-21 7:15 ` [PATCH v2 02/36] tcg: Tidy tcg_reg_alloc_op Richard Henderson
@ 2022-10-21 7:15 ` Richard Henderson
2022-10-21 7:15 ` [PATCH v2 04/36] tcg/s390x: Use register pair allocation for div and mulu2 Richard Henderson
` (32 subsequent siblings)
35 siblings, 0 replies; 43+ messages in thread
From: Richard Henderson @ 2022-10-21 7:15 UTC (permalink / raw)
To: qemu-devel
There are several instances where we need to be able to
allocate a pair of registers to related inputs/outputs.
Add 'p' and 'm' register constraints for this, in order to
be able to allocate the even/odd register first or second.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
include/tcg/tcg.h | 2 +
tcg/tcg.c | 469 ++++++++++++++++++++++++++++++++++++++--------
2 files changed, 393 insertions(+), 78 deletions(-)
diff --git a/include/tcg/tcg.h b/include/tcg/tcg.h
index d84bae6e3f..5c2254ce9f 100644
--- a/include/tcg/tcg.h
+++ b/include/tcg/tcg.h
@@ -951,6 +951,8 @@ typedef struct TCGArgConstraint {
unsigned ct : 16;
unsigned alias_index : 4;
unsigned sort_index : 4;
+ unsigned pair_index : 4;
+ unsigned pair : 2; /* 0: none, 1: first, 2: second, 3: second alias */
bool oalias : 1;
bool ialias : 1;
bool newreg : 1;
diff --git a/tcg/tcg.c b/tcg/tcg.c
index 660d5eb098..e9ff3c92e5 100644
--- a/tcg/tcg.c
+++ b/tcg/tcg.c
@@ -1985,15 +1985,32 @@ static void tcg_dump_ops(TCGContext *s, FILE *f, bool have_prefs)
static int get_constraint_priority(const TCGOpDef *def, int k)
{
const TCGArgConstraint *arg_ct = &def->args_ct[k];
- int n;
+ int n = ctpop64(arg_ct->regs);
- if (arg_ct->oalias) {
- /* an alias is equivalent to a single register */
- n = 1;
- } else {
- n = ctpop64(arg_ct->regs);
+ /*
+ * Sort constraints of a single register first, which includes output
+ * aliases (which must exactly match the input already allocated).
+ */
+ if (n == 1 || arg_ct->oalias) {
+ return INT_MAX;
}
- return TCG_TARGET_NB_REGS - n + 1;
+
+ /*
+ * Sort register pairs next, first then second immediately after.
+ * Arbitrarily sort multiple pairs by the index of the first reg;
+ * there shouldn't be many pairs.
+ */
+ switch (arg_ct->pair) {
+ case 1:
+ case 3:
+ return (k + 1) * 2;
+ case 2:
+ return (arg_ct->pair_index + 1) * 2 - 1;
+ }
+
+ /* Finally, sort by decreasing register count. */
+ assert(n > 1);
+ return -n;
}
/* sort from highest priority to lowest */
@@ -2028,7 +2045,8 @@ static void process_op_defs(TCGContext *s)
for (op = 0; op < NB_OPS; op++) {
TCGOpDef *def = &tcg_op_defs[op];
const TCGTargetOpDef *tdefs;
- int i, nb_args;
+ bool saw_alias_pair = false;
+ int i, o, i2, o2, nb_args;
if (def->flags & TCG_OPF_NOT_PRESENT) {
continue;
@@ -2050,58 +2068,175 @@ static void process_op_defs(TCGContext *s)
for (i = 0; i < nb_args; i++) {
const char *ct_str = tdefs->args_ct_str[i];
+ bool input_p = i >= def->nb_oargs;
+
/* Incomplete TCGTargetOpDef entry. */
tcg_debug_assert(ct_str != NULL);
- while (*ct_str != '\0') {
- switch(*ct_str) {
- case '0' ... '9':
- {
- int oarg = *ct_str - '0';
- tcg_debug_assert(ct_str == tdefs->args_ct_str[i]);
- tcg_debug_assert(oarg < def->nb_oargs);
- tcg_debug_assert(def->args_ct[oarg].regs != 0);
- def->args_ct[i] = def->args_ct[oarg];
- /* The output sets oalias. */
- def->args_ct[oarg].oalias = true;
- def->args_ct[oarg].alias_index = i;
- /* The input sets ialias. */
- def->args_ct[i].ialias = true;
- def->args_ct[i].alias_index = oarg;
- }
- ct_str++;
- break;
- case '&':
- def->args_ct[i].newreg = true;
- ct_str++;
- break;
+ switch (*ct_str) {
+ case '0' ... '9':
+ o = *ct_str - '0';
+ tcg_debug_assert(input_p);
+ tcg_debug_assert(o < def->nb_oargs);
+ tcg_debug_assert(def->args_ct[o].regs != 0);
+ tcg_debug_assert(!def->args_ct[o].oalias);
+ def->args_ct[i] = def->args_ct[o];
+ /* The output sets oalias. */
+ def->args_ct[o].oalias = 1;
+ def->args_ct[o].alias_index = i;
+ /* The input sets ialias. */
+ def->args_ct[i].ialias = 1;
+ def->args_ct[i].alias_index = o;
+ if (def->args_ct[i].pair) {
+ saw_alias_pair = true;
+ }
+ tcg_debug_assert(ct_str[1] == '\0');
+ continue;
+
+ case '&':
+ tcg_debug_assert(!input_p);
+ def->args_ct[i].newreg = true;
+ ct_str++;
+ break;
+
+ case 'p': /* plus */
+ /* Allocate to the register after the previous. */
+ tcg_debug_assert(i > (input_p ? def->nb_oargs : 0));
+ o = i - 1;
+ tcg_debug_assert(!def->args_ct[o].pair);
+ tcg_debug_assert(!def->args_ct[o].ct);
+ def->args_ct[i] = (TCGArgConstraint){
+ .pair = 2,
+ .pair_index = o,
+ .regs = def->args_ct[o].regs << 1,
+ };
+ def->args_ct[o].pair = 1;
+ def->args_ct[o].pair_index = i;
+ tcg_debug_assert(ct_str[1] == '\0');
+ continue;
+
+ case 'm': /* minus */
+ /* Allocate to the register before the previous. */
+ tcg_debug_assert(i > (input_p ? def->nb_oargs : 0));
+ o = i - 1;
+ tcg_debug_assert(!def->args_ct[o].pair);
+ tcg_debug_assert(!def->args_ct[o].ct);
+ def->args_ct[i] = (TCGArgConstraint){
+ .pair = 1,
+ .pair_index = o,
+ .regs = def->args_ct[o].regs >> 1,
+ };
+ def->args_ct[o].pair = 2;
+ def->args_ct[o].pair_index = i;
+ tcg_debug_assert(ct_str[1] == '\0');
+ continue;
+ }
+
+ do {
+ switch (*ct_str) {
case 'i':
def->args_ct[i].ct |= TCG_CT_CONST;
- ct_str++;
break;
/* Include all of the target-specific constraints. */
#undef CONST
#define CONST(CASE, MASK) \
- case CASE: def->args_ct[i].ct |= MASK; ct_str++; break;
+ case CASE: def->args_ct[i].ct |= MASK; break;
#define REGS(CASE, MASK) \
- case CASE: def->args_ct[i].regs |= MASK; ct_str++; break;
+ case CASE: def->args_ct[i].regs |= MASK; break;
#include "tcg-target-con-str.h"
#undef REGS
#undef CONST
default:
+ case '0' ... '9':
+ case '&':
+ case 'p':
+ case 'm':
/* Typo in TCGTargetOpDef constraint. */
g_assert_not_reached();
}
- }
+ } while (*++ct_str != '\0');
}
/* TCGTargetOpDef entry with too much information? */
tcg_debug_assert(i == TCG_MAX_OP_ARGS || tdefs->args_ct_str[i] == NULL);
+ /*
+ * Fix up output pairs that are aliased with inputs.
+ * When we created the alias, we copied pair from the output.
+ * There are three cases:
+ * (1a) Pairs of inputs alias pairs of outputs.
+ * (1b) One input aliases the first of a pair of outputs.
+ * (2) One input aliases the second of a pair of outputs.
+ *
+ * Case 1a is handled by making sure that the pair_index'es are
+ * properly updated so that they appear the same as a pair of inputs.
+ *
+ * Case 1b is handled by setting the pair_index of the input to
+ * itself, simply so it doesn't point to an unrelated argument.
+ * Since we don't encounter the "second" during the input allocation
+ * phase, nothing happens with the second half of the input pair.
+ *
+ * Case 2 is handled by setting the second input to pair=3, the
+ * first output to pair=3, and the pair_index'es to match.
+ */
+ if (saw_alias_pair) {
+ for (i = def->nb_oargs; i < nb_args; i++) {
+ /*
+ * Since [0-9pm] must be alone in the constraint string,
+ * the only way they can both be set is if the pair comes
+ * from the output alias.
+ */
+ if (!def->args_ct[i].ialias) {
+ continue;
+ }
+ switch (def->args_ct[i].pair) {
+ case 0:
+ break;
+ case 1:
+ o = def->args_ct[i].alias_index;
+ o2 = def->args_ct[o].pair_index;
+ tcg_debug_assert(def->args_ct[o].pair == 1);
+ tcg_debug_assert(def->args_ct[o2].pair == 2);
+ if (def->args_ct[o2].oalias) {
+ /* Case 1a */
+ i2 = def->args_ct[o2].alias_index;
+ tcg_debug_assert(def->args_ct[i2].pair == 2);
+ def->args_ct[i2].pair_index = i;
+ def->args_ct[i].pair_index = i2;
+ } else {
+ /* Case 1b */
+ def->args_ct[i].pair_index = i;
+ }
+ break;
+ case 2:
+ o = def->args_ct[i].alias_index;
+ o2 = def->args_ct[o].pair_index;
+ tcg_debug_assert(def->args_ct[o].pair == 2);
+ tcg_debug_assert(def->args_ct[o2].pair == 1);
+ if (def->args_ct[o2].oalias) {
+ /* Case 1a */
+ i2 = def->args_ct[o2].alias_index;
+ tcg_debug_assert(def->args_ct[i2].pair == 1);
+ def->args_ct[i2].pair_index = i;
+ def->args_ct[i].pair_index = i2;
+ } else {
+ /* Case 2 */
+ def->args_ct[i].pair = 3;
+ def->args_ct[o2].pair = 3;
+ def->args_ct[i].pair_index = o2;
+ def->args_ct[o2].pair_index = i;
+ }
+ break;
+ default:
+ g_assert_not_reached();
+ }
+ }
+ }
+
/* sort the constraints (XXX: this is just an heuristic) */
sort_constraints(def, 0, def->nb_oargs);
sort_constraints(def, def->nb_oargs, def->nb_iargs);
@@ -3197,6 +3332,52 @@ static TCGReg tcg_reg_alloc(TCGContext *s, TCGRegSet required_regs,
tcg_abort();
}
+static TCGReg tcg_reg_alloc_pair(TCGContext *s, TCGRegSet required_regs,
+ TCGRegSet allocated_regs,
+ TCGRegSet preferred_regs, bool rev)
+{
+ int i, j, k, fmin, n = ARRAY_SIZE(tcg_target_reg_alloc_order);
+ TCGRegSet reg_ct[2];
+ const int *order;
+
+ /* Ensure that if I is not in allocated_regs, I+1 is not either. */
+ reg_ct[1] = required_regs & ~(allocated_regs | (allocated_regs >> 1));
+ tcg_debug_assert(reg_ct[1] != 0);
+ reg_ct[0] = reg_ct[1] & preferred_regs;
+
+ order = rev ? indirect_reg_alloc_order : tcg_target_reg_alloc_order;
+
+ /*
+ * Skip the preferred_regs option if it cannot be satisfied,
+ * or if the preference made no difference.
+ */
+ k = reg_ct[0] == 0 || reg_ct[0] == reg_ct[1];
+
+ /*
+ * Minimize the number of flushes by looking for 2 free registers first,
+ * then a single flush, then two flushes.
+ */
+ for (fmin = 2; fmin >= 0; fmin--) {
+ for (j = k; j < 2; j++) {
+ TCGRegSet set = reg_ct[j];
+
+ for (i = 0; i < n; i++) {
+ TCGReg reg = order[i];
+
+ if (tcg_regset_test_reg(set, reg)) {
+ int f = !s->reg_to_temp[reg] + !s->reg_to_temp[reg + 1];
+ if (f >= fmin) {
+ tcg_reg_free(s, reg, allocated_regs);
+ tcg_reg_free(s, reg + 1, allocated_regs);
+ return reg;
+ }
+ }
+ }
+ }
+ }
+ tcg_abort();
+}
+
/* Make sure the temporary is in a register. If needed, allocate the register
from DESIRED while avoiding ALLOCATED. */
static void temp_load(TCGContext *s, TCGTemp *ts, TCGRegSet desired_regs,
@@ -3606,8 +3787,10 @@ static void tcg_reg_alloc_op(TCGContext *s, const TCGOp *op)
/* satisfy input constraints */
for (k = 0; k < nb_iargs; k++) {
- TCGRegSet i_preferred_regs;
- bool allocate_new_reg;
+ TCGRegSet i_preferred_regs, i_required_regs;
+ bool allocate_new_reg, copyto_new_reg;
+ TCGTemp *ts2;
+ int i1, i2;
i = def->args_ct[nb_oargs + k].sort_index;
arg = op->args[i];
@@ -3624,43 +3807,142 @@ static void tcg_reg_alloc_op(TCGContext *s, const TCGOp *op)
reg = ts->reg;
i_preferred_regs = 0;
+ i_required_regs = arg_ct->regs;
allocate_new_reg = false;
+ copyto_new_reg = false;
- if (arg_ct->ialias) {
+ switch (arg_ct->pair) {
+ case 0: /* not paired */
+ if (arg_ct->ialias) {
+ i_preferred_regs = op->output_pref[arg_ct->alias_index];
+
+ /*
+ * If the input is readonly, then it cannot also be an
+ * output and aliased to itself. If the input is not
+ * dead after the instruction, we must allocate a new
+ * register and move it.
+ */
+ if (temp_readonly(ts) || !IS_DEAD_ARG(i)) {
+ allocate_new_reg = true;
+ } else if (ts->val_type == TEMP_VAL_REG) {
+ /*
+ * Check if the current register has already been
+ * allocated for another input.
+ */
+ allocate_new_reg =
+ tcg_regset_test_reg(i_allocated_regs, reg);
+ }
+ }
+ if (!allocate_new_reg) {
+ temp_load(s, ts, i_required_regs, i_allocated_regs,
+ i_preferred_regs);
+ reg = ts->reg;
+ allocate_new_reg = !tcg_regset_test_reg(i_required_regs, reg);
+ }
+ if (allocate_new_reg) {
+ /*
+ * Allocate a new register matching the constraint
+ * and move the temporary register into it.
+ */
+ temp_load(s, ts, tcg_target_available_regs[ts->type],
+ i_allocated_regs, 0);
+ reg = tcg_reg_alloc(s, i_required_regs, i_allocated_regs,
+ i_preferred_regs, ts->indirect_base);
+ copyto_new_reg = true;
+ }
+ break;
+
+ case 1:
+ /* First of an input pair; if i1 == i2, the second is an output. */
+ i1 = i;
+ i2 = arg_ct->pair_index;
+
+ /*
+ * It is easier to default to allocating a new pair
+ * and to identify a few cases where it's not required.
+ */
+ allocate_new_reg = true;
+ ts2 = i1 != i2 ? arg_temp(op->args[i2]) : NULL;
+
+ if (arg_ct->ialias) {
+ i_preferred_regs = op->output_pref[arg_ct->alias_index];
+ if (IS_DEAD_ARG(i1) &&
+ IS_DEAD_ARG(i2) &&
+ !temp_readonly(ts) &&
+ ts->val_type == TEMP_VAL_REG &&
+ ts->reg < TCG_TARGET_NB_REGS - 1 &&
+ tcg_regset_test_reg(i_required_regs, reg) &&
+ !tcg_regset_test_reg(i_allocated_regs, reg) &&
+ !tcg_regset_test_reg(i_allocated_regs, reg + 1) &&
+ (ts2
+ ? !temp_readonly(ts2) &&
+ ts2->val_type == TEMP_VAL_REG &&
+ ts2->reg == reg + 1
+ : s->reg_to_temp[reg + 1] == NULL)) {
+ allocate_new_reg = false;
+ }
+ } else {
+ /* Without aliasing, the pair must also be an input. */
+ tcg_debug_assert(ts2);
+ if (ts->val_type == TEMP_VAL_REG &&
+ ts2->val_type == TEMP_VAL_REG &&
+ ts2->reg == reg + 1 &&
+ tcg_regset_test_reg(i_required_regs, reg)) {
+ allocate_new_reg = false;
+ }
+ }
+ if (allocate_new_reg) {
+ reg = tcg_reg_alloc_pair(s, i_required_regs, i_allocated_regs,
+ 0, ts->indirect_base);
+ if (ts->val_type != TEMP_VAL_REG) {
+ temp_load(s, ts, (TCGRegSet)1 << reg, i_allocated_regs, 0);
+ } else {
+ copyto_new_reg = true;
+ }
+ }
+ break;
+
+ case 2: /* pair second */
+ reg = new_args[arg_ct->pair_index] + 1;
+ i_required_regs = (TCGRegSet)1 << reg;
+ temp_load(s, ts, i_required_regs, i_allocated_regs, 0);
+ allocate_new_reg = ts->reg != reg;
+ break;
+
+ case 3: /* ialias with second output, no first input */
+ tcg_debug_assert(arg_ct->ialias);
i_preferred_regs = op->output_pref[arg_ct->alias_index];
- /*
- * If the input is readonly, then it cannot also be an
- * output and aliased to itself. If the input is not
- * dead after the instruction, we must allocate a new
- * register and move it.
- */
- if (temp_readonly(ts) || !IS_DEAD_ARG(i)) {
- allocate_new_reg = true;
- } else if (ts->val_type == TEMP_VAL_REG) {
- /*
- * Check if the current register has already been
- * allocated for another input.
- */
- allocate_new_reg = tcg_regset_test_reg(i_allocated_regs, reg);
+ allocate_new_reg = true;
+ if (IS_DEAD_ARG(i) &&
+ ts->val_type == TEMP_VAL_REG &&
+ reg > 0 &&
+ !s->reg_to_temp[reg - 1] &&
+ tcg_regset_test_reg(i_required_regs, reg) &&
+ !tcg_regset_test_reg(i_allocated_regs, reg) &&
+ !tcg_regset_test_reg(i_allocated_regs, reg - 1)) {
+ tcg_regset_set_reg(i_allocated_regs, reg - 1);
+ allocate_new_reg = false;
}
+ if (allocate_new_reg) {
+ reg = tcg_reg_alloc_pair(s, i_required_regs >> 1,
+ i_allocated_regs, 0,
+ ts->indirect_base);
+ tcg_regset_set_reg(i_allocated_regs, reg);
+ reg += 1;
+ if (ts->val_type != TEMP_VAL_REG) {
+ temp_load(s, ts, (TCGRegSet)1 << reg, i_allocated_regs, 0);
+ } else {
+ copyto_new_reg = true;
+ }
+ }
+ break;
+
+ default:
+ g_assert_not_reached();
}
- if (!allocate_new_reg) {
- temp_load(s, ts, arg_ct->regs, i_allocated_regs, i_preferred_regs);
- reg = ts->reg;
- allocate_new_reg = !tcg_regset_test_reg(arg_ct->regs, reg);
- }
-
- if (allocate_new_reg) {
- /*
- * Allocate a new register matching the constraint
- * and move the temporary register into it.
- */
- temp_load(s, ts, tcg_target_available_regs[ts->type],
- i_allocated_regs, 0);
- reg = tcg_reg_alloc(s, arg_ct->regs, i_allocated_regs,
- i_preferred_regs, ts->indirect_base);
+ if (copyto_new_reg) {
if (!tcg_out_mov(s, ts->type, reg, ts->reg)) {
/*
* Cross register class move not supported. Sync the
@@ -3675,7 +3957,7 @@ static void tcg_reg_alloc_op(TCGContext *s, const TCGOp *op)
const_args[i] = 0;
tcg_regset_set_reg(i_allocated_regs, reg);
}
-
+
/* mark dead temporaries and free the associated registers */
for (i = nb_oargs; i < nb_oargs + nb_iargs; i++) {
if (IS_DEAD_ARG(i)) {
@@ -3703,7 +3985,7 @@ static void tcg_reg_alloc_op(TCGContext *s, const TCGOp *op)
}
/* satisfy the output constraints */
- for(k = 0; k < nb_oargs; k++) {
+ for (k = 0; k < nb_oargs; k++) {
i = def->args_ct[k].sort_index;
arg = op->args[i];
arg_ct = &def->args_ct[i];
@@ -3712,15 +3994,46 @@ static void tcg_reg_alloc_op(TCGContext *s, const TCGOp *op)
/* ENV should not be modified. */
tcg_debug_assert(!temp_readonly(ts));
- if (arg_ct->oalias && !const_args[arg_ct->alias_index]) {
- reg = new_args[arg_ct->alias_index];
- } else if (arg_ct->newreg) {
- reg = tcg_reg_alloc(s, arg_ct->regs,
- i_allocated_regs | o_allocated_regs,
- op->output_pref[k], ts->indirect_base);
- } else {
- reg = tcg_reg_alloc(s, arg_ct->regs, o_allocated_regs,
- op->output_pref[k], ts->indirect_base);
+ switch (arg_ct->pair) {
+ case 0: /* not paired */
+ if (arg_ct->oalias && !const_args[arg_ct->alias_index]) {
+ reg = new_args[arg_ct->alias_index];
+ } else if (arg_ct->newreg) {
+ reg = tcg_reg_alloc(s, arg_ct->regs,
+ i_allocated_regs | o_allocated_regs,
+ op->output_pref[k], ts->indirect_base);
+ } else {
+ reg = tcg_reg_alloc(s, arg_ct->regs, o_allocated_regs,
+ op->output_pref[k], ts->indirect_base);
+ }
+ break;
+
+ case 1: /* first of pair */
+ tcg_debug_assert(!arg_ct->newreg);
+ if (arg_ct->oalias) {
+ reg = new_args[arg_ct->alias_index];
+ break;
+ }
+ reg = tcg_reg_alloc_pair(s, arg_ct->regs, o_allocated_regs,
+ op->output_pref[k], ts->indirect_base);
+ break;
+
+ case 2: /* second of pair */
+ tcg_debug_assert(!arg_ct->newreg);
+ if (arg_ct->oalias) {
+ reg = new_args[arg_ct->alias_index];
+ } else {
+ reg = new_args[arg_ct->pair_index] + 1;
+ }
+ break;
+
+ case 3: /* first of pair, aliasing with a second input */
+ tcg_debug_assert(!arg_ct->newreg);
+ reg = new_args[arg_ct->pair_index] - 1;
+ break;
+
+ default:
+ g_assert_not_reached();
}
tcg_regset_set_reg(o_allocated_regs, reg);
if (ts->val_type == TEMP_VAL_REG) {
@@ -3728,12 +4041,12 @@ static void tcg_reg_alloc_op(TCGContext *s, const TCGOp *op)
}
ts->val_type = TEMP_VAL_REG;
ts->reg = reg;
+ s->reg_to_temp[reg] = ts;
/*
* Temp value is modified, so the value kept in memory is
* potentially not the same.
*/
ts->mem_coherent = 0;
- s->reg_to_temp[reg] = ts;
new_args[i] = reg;
}
}
--
2.34.1
^ permalink raw reply related [flat|nested] 43+ messages in thread* [PATCH v2 04/36] tcg/s390x: Use register pair allocation for div and mulu2
2022-10-21 7:15 [PATCH v2 00/36] tcg: Support for Int128 with helpers Richard Henderson
` (2 preceding siblings ...)
2022-10-21 7:15 ` [PATCH v2 03/36] tcg: Introduce paired register allocation Richard Henderson
@ 2022-10-21 7:15 ` Richard Henderson
2022-10-21 7:15 ` [PATCH v2 05/36] tcg/arm: Use register pair allocation for qemu_{ld, st}_i64 Richard Henderson
` (31 subsequent siblings)
35 siblings, 0 replies; 43+ messages in thread
From: Richard Henderson @ 2022-10-21 7:15 UTC (permalink / raw)
To: qemu-devel
Previously we hard-coded R2 and R3.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
tcg/s390x/tcg-target-con-set.h | 4 ++--
tcg/s390x/tcg-target-con-str.h | 8 +------
tcg/s390x/tcg-target.c.inc | 43 +++++++++++++++++++++++++---------
3 files changed, 35 insertions(+), 20 deletions(-)
diff --git a/tcg/s390x/tcg-target-con-set.h b/tcg/s390x/tcg-target-con-set.h
index 426dd92e51..00ba727b70 100644
--- a/tcg/s390x/tcg-target-con-set.h
+++ b/tcg/s390x/tcg-target-con-set.h
@@ -29,8 +29,8 @@ C_O1_I2(v, v, v)
C_O1_I3(v, v, v, v)
C_O1_I4(r, r, ri, r, 0)
C_O1_I4(r, r, ri, rI, 0)
-C_O2_I2(b, a, 0, r)
-C_O2_I3(b, a, 0, 1, r)
+C_O2_I2(o, m, 0, r)
+C_O2_I3(o, m, 0, 1, r)
C_O2_I4(r, r, 0, 1, rA, r)
C_O2_I4(r, r, 0, 1, ri, r)
C_O2_I4(r, r, 0, 1, r, r)
diff --git a/tcg/s390x/tcg-target-con-str.h b/tcg/s390x/tcg-target-con-str.h
index 8bb0358ae5..76446aecae 100644
--- a/tcg/s390x/tcg-target-con-str.h
+++ b/tcg/s390x/tcg-target-con-str.h
@@ -11,13 +11,7 @@
REGS('r', ALL_GENERAL_REGS)
REGS('L', ALL_GENERAL_REGS & ~SOFTMMU_RESERVE_REGS)
REGS('v', ALL_VECTOR_REGS)
-/*
- * A (single) even/odd pair for division.
- * TODO: Add something to the register allocator to allow
- * this kind of regno+1 pairing to be done more generally.
- */
-REGS('a', 1u << TCG_REG_R2)
-REGS('b', 1u << TCG_REG_R3)
+REGS('o', 0xaaaa) /* odd numbered general regs */
/*
* Define constraint letters for constants:
diff --git a/tcg/s390x/tcg-target.c.inc b/tcg/s390x/tcg-target.c.inc
index 33becd7694..47aaba7667 100644
--- a/tcg/s390x/tcg-target.c.inc
+++ b/tcg/s390x/tcg-target.c.inc
@@ -2258,10 +2258,18 @@ static inline void tcg_out_op(TCGContext *s, TCGOpcode opc,
break;
case INDEX_op_div2_i32:
- tcg_out_insn(s, RR, DR, TCG_REG_R2, args[4]);
+ tcg_debug_assert(args[0] == args[2]);
+ tcg_debug_assert(args[1] == args[3]);
+ tcg_debug_assert((args[1] & 1) == 0);
+ tcg_debug_assert(args[0] == args[1] + 1);
+ tcg_out_insn(s, RR, DR, args[1], args[4]);
break;
case INDEX_op_divu2_i32:
- tcg_out_insn(s, RRE, DLR, TCG_REG_R2, args[4]);
+ tcg_debug_assert(args[0] == args[2]);
+ tcg_debug_assert(args[1] == args[3]);
+ tcg_debug_assert((args[1] & 1) == 0);
+ tcg_debug_assert(args[0] == args[1] + 1);
+ tcg_out_insn(s, RRE, DLR, args[1], args[4]);
break;
case INDEX_op_shl_i32:
@@ -2515,17 +2523,30 @@ static inline void tcg_out_op(TCGContext *s, TCGOpcode opc,
break;
case INDEX_op_div2_i64:
- /* ??? We get an unnecessary sign-extension of the dividend
- into R3 with this definition, but as we do in fact always
- produce both quotient and remainder using INDEX_op_div_i64
- instead requires jumping through even more hoops. */
- tcg_out_insn(s, RRE, DSGR, TCG_REG_R2, args[4]);
+ /*
+ * ??? We get an unnecessary sign-extension of the dividend
+ * into op0 with this definition, but as we do in fact always
+ * produce both quotient and remainder using INDEX_op_div_i64
+ * instead requires jumping through even more hoops.
+ */
+ tcg_debug_assert(args[0] == args[2]);
+ tcg_debug_assert(args[1] == args[3]);
+ tcg_debug_assert((args[1] & 1) == 0);
+ tcg_debug_assert(args[0] == args[1] + 1);
+ tcg_out_insn(s, RRE, DSGR, args[1], args[4]);
break;
case INDEX_op_divu2_i64:
- tcg_out_insn(s, RRE, DLGR, TCG_REG_R2, args[4]);
+ tcg_debug_assert(args[0] == args[2]);
+ tcg_debug_assert(args[1] == args[3]);
+ tcg_debug_assert((args[1] & 1) == 0);
+ tcg_debug_assert(args[0] == args[1] + 1);
+ tcg_out_insn(s, RRE, DLGR, args[1], args[4]);
break;
case INDEX_op_mulu2_i64:
- tcg_out_insn(s, RRE, MLGR, TCG_REG_R2, args[3]);
+ tcg_debug_assert(args[0] == args[2]);
+ tcg_debug_assert((args[1] & 1) == 0);
+ tcg_debug_assert(args[0] == args[1] + 1);
+ tcg_out_insn(s, RRE, MLGR, args[1], args[3]);
break;
case INDEX_op_shl_i64:
@@ -3220,10 +3241,10 @@ static TCGConstraintSetIndex tcg_target_op_def(TCGOpcode op)
case INDEX_op_div2_i64:
case INDEX_op_divu2_i32:
case INDEX_op_divu2_i64:
- return C_O2_I3(b, a, 0, 1, r);
+ return C_O2_I3(o, m, 0, 1, r);
case INDEX_op_mulu2_i64:
- return C_O2_I2(b, a, 0, r);
+ return C_O2_I2(o, m, 0, r);
case INDEX_op_add2_i32:
case INDEX_op_sub2_i32:
--
2.34.1
^ permalink raw reply related [flat|nested] 43+ messages in thread* [PATCH v2 05/36] tcg/arm: Use register pair allocation for qemu_{ld, st}_i64
2022-10-21 7:15 [PATCH v2 00/36] tcg: Support for Int128 with helpers Richard Henderson
` (3 preceding siblings ...)
2022-10-21 7:15 ` [PATCH v2 04/36] tcg/s390x: Use register pair allocation for div and mulu2 Richard Henderson
@ 2022-10-21 7:15 ` Richard Henderson
2022-10-21 7:15 ` [PATCH v2 06/36] meson: Move CONFIG_TCG_INTERPRETER to config_host Richard Henderson
` (30 subsequent siblings)
35 siblings, 0 replies; 43+ messages in thread
From: Richard Henderson @ 2022-10-21 7:15 UTC (permalink / raw)
To: qemu-devel
Although we still can't use ldrd and strd for all operations,
increase the chances by getting the register allocation correct.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
tcg/arm/tcg-target-con-set.h | 7 ++++---
tcg/arm/tcg-target-con-str.h | 2 ++
tcg/arm/tcg-target.c.inc | 28 ++++++++++++++++++----------
3 files changed, 24 insertions(+), 13 deletions(-)
diff --git a/tcg/arm/tcg-target-con-set.h b/tcg/arm/tcg-target-con-set.h
index 3685e1786a..b8849b2478 100644
--- a/tcg/arm/tcg-target-con-set.h
+++ b/tcg/arm/tcg-target-con-set.h
@@ -15,8 +15,9 @@ C_O0_I2(r, rIN)
C_O0_I2(s, s)
C_O0_I2(w, r)
C_O0_I3(s, s, s)
+C_O0_I3(S, p, s)
C_O0_I4(r, r, rI, rI)
-C_O0_I4(s, s, s, s)
+C_O0_I4(S, p, s, s)
C_O1_I1(r, l)
C_O1_I1(r, r)
C_O1_I1(w, r)
@@ -38,8 +39,8 @@ C_O1_I2(w, w, wZ)
C_O1_I3(w, w, w, w)
C_O1_I4(r, r, r, rI, rI)
C_O1_I4(r, r, rIN, rIK, 0)
-C_O2_I1(r, r, l)
-C_O2_I2(r, r, l, l)
+C_O2_I1(e, p, l)
+C_O2_I2(e, p, l, l)
C_O2_I2(r, r, r, r)
C_O2_I4(r, r, r, r, rIN, rIK)
C_O2_I4(r, r, rI, rI, rIN, rIK)
diff --git a/tcg/arm/tcg-target-con-str.h b/tcg/arm/tcg-target-con-str.h
index 8f501149e1..24b4b59feb 100644
--- a/tcg/arm/tcg-target-con-str.h
+++ b/tcg/arm/tcg-target-con-str.h
@@ -8,9 +8,11 @@
* Define constraint letters for register sets:
* REGS(letter, register_mask)
*/
+REGS('e', ALL_GENERAL_REGS & 0x5555) /* even regs */
REGS('r', ALL_GENERAL_REGS)
REGS('l', ALL_QLOAD_REGS)
REGS('s', ALL_QSTORE_REGS)
+REGS('S', ALL_QSTORE_REGS & 0x5555) /* even qstore */
REGS('w', ALL_VECTOR_REGS)
/*
diff --git a/tcg/arm/tcg-target.c.inc b/tcg/arm/tcg-target.c.inc
index 2c6c353eea..aa3a888fed 100644
--- a/tcg/arm/tcg-target.c.inc
+++ b/tcg/arm/tcg-target.c.inc
@@ -1686,9 +1686,11 @@ static void tcg_out_qemu_ld_index(TCGContext *s, MemOp opc,
tcg_out_ld32_r(s, COND_AL, datalo, addrlo, addend);
break;
case MO_UQ:
+ /* We used pair allocation for datalo, so already should be aligned. */
+ tcg_debug_assert((datalo & 1) == 0);
+ tcg_debug_assert(datahi == datalo + 1);
/* LDRD requires alignment; double-check that. */
- if (get_alignment_bits(opc) >= MO_64
- && (datalo & 1) == 0 && datahi == datalo + 1) {
+ if (get_alignment_bits(opc) >= MO_64) {
/*
* Rm (the second address op) must not overlap Rt or Rt + 1.
* Since datalo is aligned, we can simplify the test via alignment.
@@ -1742,9 +1744,11 @@ static void tcg_out_qemu_ld_direct(TCGContext *s, MemOp opc, TCGReg datalo,
tcg_out_ld32_12(s, COND_AL, datalo, addrlo, 0);
break;
case MO_UQ:
+ /* We used pair allocation for datalo, so already should be aligned. */
+ tcg_debug_assert((datalo & 1) == 0);
+ tcg_debug_assert(datahi == datalo + 1);
/* LDRD requires alignment; double-check that. */
- if (get_alignment_bits(opc) >= MO_64
- && (datalo & 1) == 0 && datahi == datalo + 1) {
+ if (get_alignment_bits(opc) >= MO_64) {
tcg_out_ldrd_8(s, COND_AL, datalo, addrlo, 0);
} else if (datalo == addrlo) {
tcg_out_ld32_12(s, COND_AL, datahi, addrlo, 4);
@@ -1826,9 +1830,11 @@ static void tcg_out_qemu_st_index(TCGContext *s, ARMCond cond, MemOp opc,
tcg_out_st32_r(s, cond, datalo, addrlo, addend);
break;
case MO_64:
+ /* We used pair allocation for datalo, so already should be aligned. */
+ tcg_debug_assert((datalo & 1) == 0);
+ tcg_debug_assert(datahi == datalo + 1);
/* STRD requires alignment; double-check that. */
- if (get_alignment_bits(opc) >= MO_64
- && (datalo & 1) == 0 && datahi == datalo + 1) {
+ if (get_alignment_bits(opc) >= MO_64) {
tcg_out_strd_r(s, cond, datalo, addrlo, addend);
} else if (scratch_addend) {
tcg_out_st32_rwb(s, cond, datalo, addend, addrlo);
@@ -1863,9 +1869,11 @@ static void tcg_out_qemu_st_direct(TCGContext *s, MemOp opc, TCGReg datalo,
tcg_out_st32_12(s, COND_AL, datalo, addrlo, 0);
break;
case MO_64:
+ /* We used pair allocation for datalo, so already should be aligned. */
+ tcg_debug_assert((datalo & 1) == 0);
+ tcg_debug_assert(datahi == datalo + 1);
/* STRD requires alignment; double-check that. */
- if (get_alignment_bits(opc) >= MO_64
- && (datalo & 1) == 0 && datahi == datalo + 1) {
+ if (get_alignment_bits(opc) >= MO_64) {
tcg_out_strd_8(s, COND_AL, datalo, addrlo, 0);
} else {
tcg_out_st32_12(s, COND_AL, datalo, addrlo, 0);
@@ -2333,11 +2341,11 @@ static TCGConstraintSetIndex tcg_target_op_def(TCGOpcode op)
case INDEX_op_qemu_ld_i32:
return TARGET_LONG_BITS == 32 ? C_O1_I1(r, l) : C_O1_I2(r, l, l);
case INDEX_op_qemu_ld_i64:
- return TARGET_LONG_BITS == 32 ? C_O2_I1(r, r, l) : C_O2_I2(r, r, l, l);
+ return TARGET_LONG_BITS == 32 ? C_O2_I1(e, p, l) : C_O2_I2(e, p, l, l);
case INDEX_op_qemu_st_i32:
return TARGET_LONG_BITS == 32 ? C_O0_I2(s, s) : C_O0_I3(s, s, s);
case INDEX_op_qemu_st_i64:
- return TARGET_LONG_BITS == 32 ? C_O0_I3(s, s, s) : C_O0_I4(s, s, s, s);
+ return TARGET_LONG_BITS == 32 ? C_O0_I3(S, p, s) : C_O0_I4(S, p, s, s);
case INDEX_op_st_vec:
return C_O0_I2(w, r);
--
2.34.1
^ permalink raw reply related [flat|nested] 43+ messages in thread* [PATCH v2 06/36] meson: Move CONFIG_TCG_INTERPRETER to config_host
2022-10-21 7:15 [PATCH v2 00/36] tcg: Support for Int128 with helpers Richard Henderson
` (4 preceding siblings ...)
2022-10-21 7:15 ` [PATCH v2 05/36] tcg/arm: Use register pair allocation for qemu_{ld, st}_i64 Richard Henderson
@ 2022-10-21 7:15 ` Richard Henderson
2022-10-21 7:15 ` [PATCH v2 07/36] tcg: Remove TCG_TARGET_STACK_GROWSUP Richard Henderson
` (29 subsequent siblings)
35 siblings, 0 replies; 43+ messages in thread
From: Richard Henderson @ 2022-10-21 7:15 UTC (permalink / raw)
To: qemu-devel; +Cc: Philippe Mathieu-Daudé
Like CONFIG_TCG, the enabled method of execution is a host property
not a guest property. This exposes the define to compile-once files.
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
meson.build | 4 +---
1 file changed, 1 insertion(+), 3 deletions(-)
diff --git a/meson.build b/meson.build
index 1ec3f72edc..64c87d45a5 100644
--- a/meson.build
+++ b/meson.build
@@ -469,6 +469,7 @@ if get_option('tcg').allowed()
endif
if get_option('tcg_interpreter')
tcg_arch = 'tci'
+ config_host += { 'CONFIG_TCG_INTERPRETER': 'y' }
elif host_arch == 'x86_64'
tcg_arch = 'i386'
elif host_arch == 'ppc64'
@@ -2507,9 +2508,6 @@ foreach target : target_dirs
if sym == 'CONFIG_TCG' or target in accelerator_targets.get(sym, [])
config_target += { sym: 'y' }
config_all += { sym: 'y' }
- if sym == 'CONFIG_TCG' and tcg_arch == 'tci'
- config_target += { 'CONFIG_TCG_INTERPRETER': 'y' }
- endif
if target in modular_tcg
config_target += { 'CONFIG_TCG_MODULAR': 'y' }
else
--
2.34.1
^ permalink raw reply related [flat|nested] 43+ messages in thread* [PATCH v2 07/36] tcg: Remove TCG_TARGET_STACK_GROWSUP
2022-10-21 7:15 [PATCH v2 00/36] tcg: Support for Int128 with helpers Richard Henderson
` (5 preceding siblings ...)
2022-10-21 7:15 ` [PATCH v2 06/36] meson: Move CONFIG_TCG_INTERPRETER to config_host Richard Henderson
@ 2022-10-21 7:15 ` Richard Henderson
2022-10-21 7:15 ` [PATCH v2 08/36] accel/tcg: Set cflags_next_tb in cpu_common_initfn Richard Henderson
` (28 subsequent siblings)
35 siblings, 0 replies; 43+ messages in thread
From: Richard Henderson @ 2022-10-21 7:15 UTC (permalink / raw)
To: qemu-devel; +Cc: Philippe Mathieu-Daudé
The hppa host code has been removed since 2013; this
should have been deleted at the same time.
Fixes: 802b5081233a ("tcg-hppa: Remove tcg backend")
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
tcg/aarch64/tcg-target.h | 1 -
tcg/arm/tcg-target.h | 1 -
tcg/tcg.c | 32 ++------------------------------
3 files changed, 2 insertions(+), 32 deletions(-)
diff --git a/tcg/aarch64/tcg-target.h b/tcg/aarch64/tcg-target.h
index 485f685bd2..e145d50fef 100644
--- a/tcg/aarch64/tcg-target.h
+++ b/tcg/aarch64/tcg-target.h
@@ -16,7 +16,6 @@
#define TCG_TARGET_INSN_UNIT_SIZE 4
#define TCG_TARGET_TLB_DISPLACEMENT_BITS 24
#define MAX_CODE_GEN_BUFFER_SIZE (2 * GiB)
-#undef TCG_TARGET_STACK_GROWSUP
typedef enum {
TCG_REG_X0, TCG_REG_X1, TCG_REG_X2, TCG_REG_X3,
diff --git a/tcg/arm/tcg-target.h b/tcg/arm/tcg-target.h
index 7e96495392..56c1ac4586 100644
--- a/tcg/arm/tcg-target.h
+++ b/tcg/arm/tcg-target.h
@@ -30,7 +30,6 @@ extern int arm_arch;
#define use_armv7_instructions (__ARM_ARCH >= 7 || arm_arch >= 7)
-#undef TCG_TARGET_STACK_GROWSUP
#define TCG_TARGET_INSN_UNIT_SIZE 4
#define TCG_TARGET_TLB_DISPLACEMENT_BITS 16
#define MAX_CODE_GEN_BUFFER_SIZE UINT32_MAX
diff --git a/tcg/tcg.c b/tcg/tcg.c
index e9ff3c92e5..6c4f949c85 100644
--- a/tcg/tcg.c
+++ b/tcg/tcg.c
@@ -1551,25 +1551,8 @@ void tcg_gen_callN(void *func, TCGTemp *ret, int nargs, TCGTemp **args)
}
if (TCG_TARGET_REG_BITS < 64 && is_64bit) {
- /*
- * If stack grows up, then we will be placing successive
- * arguments at lower addresses, which means we need to
- * reverse the order compared to how we would normally
- * treat either big or little-endian. For those arguments
- * that will wind up in registers, this still works for
- * HPPA (the only current STACK_GROWSUP target) since the
- * argument registers are *also* allocated in decreasing
- * order. If another such target is added, this logic may
- * have to get more complicated to differentiate between
- * stack arguments and register arguments.
- */
-#if HOST_BIG_ENDIAN != defined(TCG_TARGET_STACK_GROWSUP)
- op->args[pi++] = temp_arg(args[i] + 1);
- op->args[pi++] = temp_arg(args[i]);
-#else
- op->args[pi++] = temp_arg(args[i]);
- op->args[pi++] = temp_arg(args[i] + 1);
-#endif
+ op->args[pi++] = temp_arg(args[i] + HOST_BIG_ENDIAN);
+ op->args[pi++] = temp_arg(args[i] + !HOST_BIG_ENDIAN);
real_args += 2;
continue;
}
@@ -4166,12 +4149,6 @@ static bool tcg_reg_alloc_dup2(TCGContext *s, const TCGOp *op)
return true;
}
-#ifdef TCG_TARGET_STACK_GROWSUP
-#define STACK_DIR(x) (-(x))
-#else
-#define STACK_DIR(x) (x)
-#endif
-
static void tcg_reg_alloc_call(TCGContext *s, TCGOp *op)
{
const int nb_oargs = TCGOP_CALLO(op);
@@ -4211,18 +4188,13 @@ static void tcg_reg_alloc_call(TCGContext *s, TCGOp *op)
stack_offset = TCG_TARGET_CALL_STACK_OFFSET;
for (i = nb_regs; i < nb_iargs; i++) {
arg = op->args[nb_oargs + i];
-#ifdef TCG_TARGET_STACK_GROWSUP
- stack_offset -= sizeof(tcg_target_long);
-#endif
if (arg != TCG_CALL_DUMMY_ARG) {
ts = arg_temp(arg);
temp_load(s, ts, tcg_target_available_regs[ts->type],
s->reserved_regs, 0);
tcg_out_st(s, ts->type, ts->reg, TCG_REG_CALL_STACK, stack_offset);
}
-#ifndef TCG_TARGET_STACK_GROWSUP
stack_offset += sizeof(tcg_target_long);
-#endif
}
/* assign input registers */
--
2.34.1
^ permalink raw reply related [flat|nested] 43+ messages in thread* [PATCH v2 08/36] accel/tcg: Set cflags_next_tb in cpu_common_initfn
2022-10-21 7:15 [PATCH v2 00/36] tcg: Support for Int128 with helpers Richard Henderson
` (6 preceding siblings ...)
2022-10-21 7:15 ` [PATCH v2 07/36] tcg: Remove TCG_TARGET_STACK_GROWSUP Richard Henderson
@ 2022-10-21 7:15 ` Richard Henderson
2022-10-21 7:15 ` [PATCH v2 09/36] target/sparc: Avoid TCGV_{LOW,HIGH} Richard Henderson
` (27 subsequent siblings)
35 siblings, 0 replies; 43+ messages in thread
From: Richard Henderson @ 2022-10-21 7:15 UTC (permalink / raw)
To: qemu-devel
While we initialize this value in cpu_common_reset, that
isn't called during startup, so set it as well in init.
This fixes -singlestep versus the very first TB.
Fixes: 04f5b647ed07 ("accel/tcg: Handle -singlestep in curr_cflags")
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
hw/core/cpu-common.c | 1 +
1 file changed, 1 insertion(+)
diff --git a/hw/core/cpu-common.c b/hw/core/cpu-common.c
index f9fdd46b9d..6a4022eb14 100644
--- a/hw/core/cpu-common.c
+++ b/hw/core/cpu-common.c
@@ -235,6 +235,7 @@ static void cpu_common_initfn(Object *obj)
/* the default value is changed by qemu_init_vcpu() for softmmu */
cpu->nr_cores = 1;
cpu->nr_threads = 1;
+ cpu->cflags_next_tb = -1;
qemu_mutex_init(&cpu->work_mutex);
QSIMPLEQ_INIT(&cpu->work_list);
--
2.34.1
^ permalink raw reply related [flat|nested] 43+ messages in thread* [PATCH v2 09/36] target/sparc: Avoid TCGV_{LOW,HIGH}
2022-10-21 7:15 [PATCH v2 00/36] tcg: Support for Int128 with helpers Richard Henderson
` (7 preceding siblings ...)
2022-10-21 7:15 ` [PATCH v2 08/36] accel/tcg: Set cflags_next_tb in cpu_common_initfn Richard Henderson
@ 2022-10-21 7:15 ` Richard Henderson
2022-10-21 7:15 ` [PATCH v2 10/36] tcg: Move TCG_{LOW,HIGH} to tcg-internal.h Richard Henderson
` (26 subsequent siblings)
35 siblings, 0 replies; 43+ messages in thread
From: Richard Henderson @ 2022-10-21 7:15 UTC (permalink / raw)
To: qemu-devel; +Cc: Mark Cave-Ayland
Use the official extend/extract functions instead of routines
that will shortly be internal to tcg.
Cc: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/sparc/translate.c | 21 ++++-----------------
1 file changed, 4 insertions(+), 17 deletions(-)
diff --git a/target/sparc/translate.c b/target/sparc/translate.c
index 2cbbe2396a..7e9631f004 100644
--- a/target/sparc/translate.c
+++ b/target/sparc/translate.c
@@ -163,13 +163,6 @@ static inline void gen_update_fprs_dirty(DisasContext *dc, int rd)
/* floating point registers moves */
static TCGv_i32 gen_load_fpr_F(DisasContext *dc, unsigned int src)
{
-#if TCG_TARGET_REG_BITS == 32
- if (src & 1) {
- return TCGV_LOW(cpu_fpr[src / 2]);
- } else {
- return TCGV_HIGH(cpu_fpr[src / 2]);
- }
-#else
TCGv_i32 ret = get_temp_i32(dc);
if (src & 1) {
tcg_gen_extrl_i64_i32(ret, cpu_fpr[src / 2]);
@@ -177,22 +170,16 @@ static TCGv_i32 gen_load_fpr_F(DisasContext *dc, unsigned int src)
tcg_gen_extrh_i64_i32(ret, cpu_fpr[src / 2]);
}
return ret;
-#endif
}
static void gen_store_fpr_F(DisasContext *dc, unsigned int dst, TCGv_i32 v)
{
-#if TCG_TARGET_REG_BITS == 32
- if (dst & 1) {
- tcg_gen_mov_i32(TCGV_LOW(cpu_fpr[dst / 2]), v);
- } else {
- tcg_gen_mov_i32(TCGV_HIGH(cpu_fpr[dst / 2]), v);
- }
-#else
- TCGv_i64 t = (TCGv_i64)v;
+ TCGv_i64 t = tcg_temp_new_i64();
+
+ tcg_gen_extu_i32_i64(t, v);
tcg_gen_deposit_i64(cpu_fpr[dst / 2], cpu_fpr[dst / 2], t,
(dst & 1 ? 0 : 32), 32);
-#endif
+ tcg_temp_free_i64(t);
gen_update_fprs_dirty(dc, dst);
}
--
2.34.1
^ permalink raw reply related [flat|nested] 43+ messages in thread* [PATCH v2 10/36] tcg: Move TCG_{LOW,HIGH} to tcg-internal.h
2022-10-21 7:15 [PATCH v2 00/36] tcg: Support for Int128 with helpers Richard Henderson
` (8 preceding siblings ...)
2022-10-21 7:15 ` [PATCH v2 09/36] target/sparc: Avoid TCGV_{LOW,HIGH} Richard Henderson
@ 2022-10-21 7:15 ` Richard Henderson
2022-10-21 7:15 ` [PATCH v2 11/36] tcg: Add temp_subindex to TCGTemp Richard Henderson
` (25 subsequent siblings)
35 siblings, 0 replies; 43+ messages in thread
From: Richard Henderson @ 2022-10-21 7:15 UTC (permalink / raw)
To: qemu-devel; +Cc: Philippe Mathieu-Daudé
Move the error-generating fallback from tcg-op.c, and
replace "_link_error" with modern QEMU_ERROR markup.
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
include/tcg/tcg.h | 12 ------------
tcg/tcg-internal.h | 14 ++++++++++++++
tcg/tcg-op.c | 10 +---------
3 files changed, 15 insertions(+), 21 deletions(-)
diff --git a/include/tcg/tcg.h b/include/tcg/tcg.h
index 5c2254ce9f..d207bc47be 100644
--- a/include/tcg/tcg.h
+++ b/include/tcg/tcg.h
@@ -737,18 +737,6 @@ static inline TCGv_vec temp_tcgv_vec(TCGTemp *t)
return (TCGv_vec)temp_tcgv_i32(t);
}
-#if TCG_TARGET_REG_BITS == 32
-static inline TCGv_i32 TCGV_LOW(TCGv_i64 t)
-{
- return temp_tcgv_i32(tcgv_i64_temp(t));
-}
-
-static inline TCGv_i32 TCGV_HIGH(TCGv_i64 t)
-{
- return temp_tcgv_i32(tcgv_i64_temp(t) + 1);
-}
-#endif
-
static inline TCGArg tcg_get_insn_param(TCGOp *op, int arg)
{
return op->args[arg];
diff --git a/tcg/tcg-internal.h b/tcg/tcg-internal.h
index cc82088d52..a9ea27f67a 100644
--- a/tcg/tcg-internal.h
+++ b/tcg/tcg-internal.h
@@ -59,4 +59,18 @@ static inline unsigned tcg_call_flags(TCGOp *op)
return tcg_call_info(op)->flags;
}
+#if TCG_TARGET_REG_BITS == 32
+static inline TCGv_i32 TCGV_LOW(TCGv_i64 t)
+{
+ return temp_tcgv_i32(tcgv_i64_temp(t));
+}
+static inline TCGv_i32 TCGV_HIGH(TCGv_i64 t)
+{
+ return temp_tcgv_i32(tcgv_i64_temp(t) + 1);
+}
+#else
+extern TCGv_i32 TCGV_LOW(TCGv_i64) QEMU_ERROR("32-bit code path is reachable");
+extern TCGv_i32 TCGV_HIGH(TCGv_i64) QEMU_ERROR("32-bit code path is reachable");
+#endif
+
#endif /* TCG_INTERNAL_H */
diff --git a/tcg/tcg-op.c b/tcg/tcg-op.c
index 019fab00cc..3ed98ffa01 100644
--- a/tcg/tcg-op.c
+++ b/tcg/tcg-op.c
@@ -28,16 +28,8 @@
#include "tcg/tcg-op.h"
#include "tcg/tcg-mo.h"
#include "exec/plugin-gen.h"
+#include "tcg-internal.h"
-/* Reduce the number of ifdefs below. This assumes that all uses of
- TCGV_HIGH and TCGV_LOW are properly protected by a conditional that
- the compiler can eliminate. */
-#if TCG_TARGET_REG_BITS == 64
-extern TCGv_i32 TCGV_LOW_link_error(TCGv_i64);
-extern TCGv_i32 TCGV_HIGH_link_error(TCGv_i64);
-#define TCGV_LOW TCGV_LOW_link_error
-#define TCGV_HIGH TCGV_HIGH_link_error
-#endif
void tcg_gen_op1(TCGOpcode opc, TCGArg a1)
{
--
2.34.1
^ permalink raw reply related [flat|nested] 43+ messages in thread* [PATCH v2 11/36] tcg: Add temp_subindex to TCGTemp
2022-10-21 7:15 [PATCH v2 00/36] tcg: Support for Int128 with helpers Richard Henderson
` (9 preceding siblings ...)
2022-10-21 7:15 ` [PATCH v2 10/36] tcg: Move TCG_{LOW,HIGH} to tcg-internal.h Richard Henderson
@ 2022-10-21 7:15 ` Richard Henderson
2022-10-21 7:15 ` [PATCH v2 12/36] tcg: Simplify calls to temp_sync vs mem_coherent Richard Henderson
` (24 subsequent siblings)
35 siblings, 0 replies; 43+ messages in thread
From: Richard Henderson @ 2022-10-21 7:15 UTC (permalink / raw)
To: qemu-devel; +Cc: Philippe Mathieu-Daudé
Record the location of a TCGTemp within a larger object.
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
include/tcg/tcg.h | 1 +
tcg/tcg.c | 3 +++
2 files changed, 4 insertions(+)
diff --git a/include/tcg/tcg.h b/include/tcg/tcg.h
index d207bc47be..afa18986b1 100644
--- a/include/tcg/tcg.h
+++ b/include/tcg/tcg.h
@@ -456,6 +456,7 @@ typedef struct TCGTemp {
unsigned int mem_coherent:1;
unsigned int mem_allocated:1;
unsigned int temp_allocated:1;
+ unsigned int temp_subindex:1;
int64_t val;
struct TCGTemp *mem_base;
diff --git a/tcg/tcg.c b/tcg/tcg.c
index 6c4f949c85..135d9a9a0a 100644
--- a/tcg/tcg.c
+++ b/tcg/tcg.c
@@ -927,6 +927,7 @@ TCGTemp *tcg_global_mem_new_internal(TCGType type, TCGv_ptr base,
ts2->mem_allocated = 1;
ts2->mem_base = base_ts;
ts2->mem_offset = offset + (1 - bigendian) * 4;
+ ts2->temp_subindex = 1;
pstrcpy(buf, sizeof(buf), name);
pstrcat(buf, sizeof(buf), "_1");
ts2->name = strdup(buf);
@@ -973,6 +974,7 @@ TCGTemp *tcg_temp_new_internal(TCGType type, bool temp_local)
ts2->base_type = TCG_TYPE_I64;
ts2->type = TCG_TYPE_I32;
ts2->temp_allocated = 1;
+ ts2->temp_subindex = 1;
ts2->kind = kind;
} else {
ts->base_type = type;
@@ -1091,6 +1093,7 @@ TCGTemp *tcg_constant_internal(TCGType type, int64_t val)
ts2->type = TCG_TYPE_I32;
ts2->kind = TEMP_CONST;
ts2->temp_allocated = 1;
+ ts2->temp_subindex = 1;
ts2->val = val >> 32;
} else {
ts->base_type = type;
--
2.34.1
^ permalink raw reply related [flat|nested] 43+ messages in thread* [PATCH v2 12/36] tcg: Simplify calls to temp_sync vs mem_coherent
2022-10-21 7:15 [PATCH v2 00/36] tcg: Support for Int128 with helpers Richard Henderson
` (10 preceding siblings ...)
2022-10-21 7:15 ` [PATCH v2 11/36] tcg: Add temp_subindex to TCGTemp Richard Henderson
@ 2022-10-21 7:15 ` Richard Henderson
2022-10-21 7:15 ` [PATCH v2 13/36] tcg: Allocate TCGTemp pairs in host memory order Richard Henderson
` (23 subsequent siblings)
35 siblings, 0 replies; 43+ messages in thread
From: Richard Henderson @ 2022-10-21 7:15 UTC (permalink / raw)
To: qemu-devel
The first thing that temp_sync does is check mem_coherent,
so there's no need for the caller to do so.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
tcg/tcg.c | 8 ++------
1 file changed, 2 insertions(+), 6 deletions(-)
diff --git a/tcg/tcg.c b/tcg/tcg.c
index 135d9a9a0a..9fd58e46b9 100644
--- a/tcg/tcg.c
+++ b/tcg/tcg.c
@@ -4117,12 +4117,8 @@ static bool tcg_reg_alloc_dup2(TCGContext *s, const TCGOp *op)
/* If the two inputs form one 64-bit value, try dupm_vec. */
if (itsl + 1 == itsh && itsl->base_type == TCG_TYPE_I64) {
- if (!itsl->mem_coherent) {
- temp_sync(s, itsl, s->reserved_regs, 0, 0);
- }
- if (!itsh->mem_coherent) {
- temp_sync(s, itsh, s->reserved_regs, 0, 0);
- }
+ temp_sync(s, itsl, s->reserved_regs, 0, 0);
+ temp_sync(s, itsh, s->reserved_regs, 0, 0);
#if HOST_BIG_ENDIAN
TCGTemp *its = itsh;
#else
--
2.34.1
^ permalink raw reply related [flat|nested] 43+ messages in thread* [PATCH v2 13/36] tcg: Allocate TCGTemp pairs in host memory order
2022-10-21 7:15 [PATCH v2 00/36] tcg: Support for Int128 with helpers Richard Henderson
` (11 preceding siblings ...)
2022-10-21 7:15 ` [PATCH v2 12/36] tcg: Simplify calls to temp_sync vs mem_coherent Richard Henderson
@ 2022-10-21 7:15 ` Richard Henderson
2022-10-21 7:15 ` [PATCH v2 14/36] tcg: Move TCG_TYPE_COUNT outside enum Richard Henderson
` (22 subsequent siblings)
35 siblings, 0 replies; 43+ messages in thread
From: Richard Henderson @ 2022-10-21 7:15 UTC (permalink / raw)
To: qemu-devel
Allocate the first of a pair at the lower address, and the
second of a pair at the higher address. This will make it
easier to find the beginning of the larger memory block.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
tcg/tcg-internal.h | 4 ++--
tcg/tcg.c | 58 ++++++++++++++++++++++------------------------
2 files changed, 30 insertions(+), 32 deletions(-)
diff --git a/tcg/tcg-internal.h b/tcg/tcg-internal.h
index a9ea27f67a..2c06b5116a 100644
--- a/tcg/tcg-internal.h
+++ b/tcg/tcg-internal.h
@@ -62,11 +62,11 @@ static inline unsigned tcg_call_flags(TCGOp *op)
#if TCG_TARGET_REG_BITS == 32
static inline TCGv_i32 TCGV_LOW(TCGv_i64 t)
{
- return temp_tcgv_i32(tcgv_i64_temp(t));
+ return temp_tcgv_i32(tcgv_i64_temp(t) + HOST_BIG_ENDIAN);
}
static inline TCGv_i32 TCGV_HIGH(TCGv_i64 t)
{
- return temp_tcgv_i32(tcgv_i64_temp(t) + 1);
+ return temp_tcgv_i32(tcgv_i64_temp(t) + !HOST_BIG_ENDIAN);
}
#else
extern TCGv_i32 TCGV_LOW(TCGv_i64) QEMU_ERROR("32-bit code path is reachable");
diff --git a/tcg/tcg.c b/tcg/tcg.c
index 9fd58e46b9..73d8b33b06 100644
--- a/tcg/tcg.c
+++ b/tcg/tcg.c
@@ -886,10 +886,7 @@ TCGTemp *tcg_global_mem_new_internal(TCGType type, TCGv_ptr base,
TCGContext *s = tcg_ctx;
TCGTemp *base_ts = tcgv_ptr_temp(base);
TCGTemp *ts = tcg_global_alloc(s);
- int indirect_reg = 0, bigendian = 0;
-#if HOST_BIG_ENDIAN
- bigendian = 1;
-#endif
+ int indirect_reg = 0;
switch (base_ts->kind) {
case TEMP_FIXED:
@@ -915,7 +912,7 @@ TCGTemp *tcg_global_mem_new_internal(TCGType type, TCGv_ptr base,
ts->indirect_reg = indirect_reg;
ts->mem_allocated = 1;
ts->mem_base = base_ts;
- ts->mem_offset = offset + bigendian * 4;
+ ts->mem_offset = offset;
pstrcpy(buf, sizeof(buf), name);
pstrcat(buf, sizeof(buf), "_0");
ts->name = strdup(buf);
@@ -926,7 +923,7 @@ TCGTemp *tcg_global_mem_new_internal(TCGType type, TCGv_ptr base,
ts2->indirect_reg = indirect_reg;
ts2->mem_allocated = 1;
ts2->mem_base = base_ts;
- ts2->mem_offset = offset + (1 - bigendian) * 4;
+ ts2->mem_offset = offset + 4;
ts2->temp_subindex = 1;
pstrcpy(buf, sizeof(buf), name);
pstrcat(buf, sizeof(buf), "_1");
@@ -1072,37 +1069,43 @@ TCGTemp *tcg_constant_internal(TCGType type, int64_t val)
ts = g_hash_table_lookup(h, &val);
if (ts == NULL) {
+ int64_t *val_ptr;
+
ts = tcg_temp_alloc(s);
if (TCG_TARGET_REG_BITS == 32 && type == TCG_TYPE_I64) {
TCGTemp *ts2 = tcg_temp_alloc(s);
+ tcg_debug_assert(ts2 == ts + 1);
+
ts->base_type = TCG_TYPE_I64;
ts->type = TCG_TYPE_I32;
ts->kind = TEMP_CONST;
ts->temp_allocated = 1;
- /*
- * Retain the full value of the 64-bit constant in the low
- * part, so that the hash table works. Actual uses will
- * truncate the value to the low part.
- */
- ts->val = val;
- tcg_debug_assert(ts2 == ts + 1);
ts2->base_type = TCG_TYPE_I64;
ts2->type = TCG_TYPE_I32;
ts2->kind = TEMP_CONST;
ts2->temp_allocated = 1;
ts2->temp_subindex = 1;
- ts2->val = val >> 32;
+
+ /*
+ * Retain the full value of the 64-bit constant in the low
+ * part, so that the hash table works. Actual uses will
+ * truncate the value to the low part.
+ */
+ ts[HOST_BIG_ENDIAN].val = val;
+ ts[!HOST_BIG_ENDIAN].val = val >> 32;
+ val_ptr = &ts[HOST_BIG_ENDIAN].val;
} else {
ts->base_type = type;
ts->type = type;
ts->kind = TEMP_CONST;
ts->temp_allocated = 1;
ts->val = val;
+ val_ptr = &ts->val;
}
- g_hash_table_insert(h, &ts->val, ts);
+ g_hash_table_insert(h, val_ptr, ts);
}
return ts;
@@ -1514,13 +1517,8 @@ void tcg_gen_callN(void *func, TCGTemp *ret, int nargs, TCGTemp **args)
pi = 0;
if (ret != NULL) {
if (TCG_TARGET_REG_BITS < 64 && (typemask & 6) == dh_typecode_i64) {
-#if HOST_BIG_ENDIAN
- op->args[pi++] = temp_arg(ret + 1);
- op->args[pi++] = temp_arg(ret);
-#else
op->args[pi++] = temp_arg(ret);
op->args[pi++] = temp_arg(ret + 1);
-#endif
nb_rets = 2;
} else {
op->args[pi++] = temp_arg(ret);
@@ -1554,8 +1552,8 @@ void tcg_gen_callN(void *func, TCGTemp *ret, int nargs, TCGTemp **args)
}
if (TCG_TARGET_REG_BITS < 64 && is_64bit) {
- op->args[pi++] = temp_arg(args[i] + HOST_BIG_ENDIAN);
- op->args[pi++] = temp_arg(args[i] + !HOST_BIG_ENDIAN);
+ op->args[pi++] = temp_arg(args[i]);
+ op->args[pi++] = temp_arg(args[i] + 1);
real_args += 2;
continue;
}
@@ -4116,14 +4114,14 @@ static bool tcg_reg_alloc_dup2(TCGContext *s, const TCGOp *op)
}
/* If the two inputs form one 64-bit value, try dupm_vec. */
- if (itsl + 1 == itsh && itsl->base_type == TCG_TYPE_I64) {
- temp_sync(s, itsl, s->reserved_regs, 0, 0);
- temp_sync(s, itsh, s->reserved_regs, 0, 0);
-#if HOST_BIG_ENDIAN
- TCGTemp *its = itsh;
-#else
- TCGTemp *its = itsl;
-#endif
+ if (itsl->temp_subindex == HOST_BIG_ENDIAN &&
+ itsh->temp_subindex == !HOST_BIG_ENDIAN &&
+ itsl == itsh + (HOST_BIG_ENDIAN ? 1 : -1)) {
+ TCGTemp *its = itsl - HOST_BIG_ENDIAN;
+
+ temp_sync(s, its + 0, s->reserved_regs, 0, 0);
+ temp_sync(s, its + 1, s->reserved_regs, 0, 0);
+
if (tcg_out_dupm_vec(s, vtype, MO_64, ots->reg,
its->mem_base->reg, its->mem_offset)) {
goto done;
--
2.34.1
^ permalink raw reply related [flat|nested] 43+ messages in thread* [PATCH v2 14/36] tcg: Move TCG_TYPE_COUNT outside enum
2022-10-21 7:15 [PATCH v2 00/36] tcg: Support for Int128 with helpers Richard Henderson
` (12 preceding siblings ...)
2022-10-21 7:15 ` [PATCH v2 13/36] tcg: Allocate TCGTemp pairs in host memory order Richard Henderson
@ 2022-10-21 7:15 ` Richard Henderson
2022-10-21 7:15 ` [PATCH v2 15/36] tcg: Introduce tcg_type_size Richard Henderson
` (21 subsequent siblings)
35 siblings, 0 replies; 43+ messages in thread
From: Richard Henderson @ 2022-10-21 7:15 UTC (permalink / raw)
To: qemu-devel; +Cc: Philippe Mathieu-Daudé
The count is not itself an enumerator. Move it outside to
prevent the compiler from considering it with -Wswitch-enum.
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
include/tcg/tcg.h | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/include/tcg/tcg.h b/include/tcg/tcg.h
index afa18986b1..f2da340bb9 100644
--- a/include/tcg/tcg.h
+++ b/include/tcg/tcg.h
@@ -294,7 +294,8 @@ typedef enum TCGType {
TCG_TYPE_V128,
TCG_TYPE_V256,
- TCG_TYPE_COUNT, /* number of different types */
+ /* Number of different types (integer not enum) */
+#define TCG_TYPE_COUNT (TCG_TYPE_V256 + 1)
/* An alias for the size of the host register. */
#if TCG_TARGET_REG_BITS == 32
--
2.34.1
^ permalink raw reply related [flat|nested] 43+ messages in thread* [PATCH v2 15/36] tcg: Introduce tcg_type_size
2022-10-21 7:15 [PATCH v2 00/36] tcg: Support for Int128 with helpers Richard Henderson
` (13 preceding siblings ...)
2022-10-21 7:15 ` [PATCH v2 14/36] tcg: Move TCG_TYPE_COUNT outside enum Richard Henderson
@ 2022-10-21 7:15 ` Richard Henderson
2022-10-21 7:15 ` [PATCH v2 16/36] tcg: Introduce TCGCallReturnKind and TCGCallArgumentKind Richard Henderson
` (20 subsequent siblings)
35 siblings, 0 replies; 43+ messages in thread
From: Richard Henderson @ 2022-10-21 7:15 UTC (permalink / raw)
To: qemu-devel
Add a helper function for computing the size of a type.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
include/tcg/tcg.h | 16 ++++++++++++++++
tcg/tcg.c | 26 ++++++++++++--------------
2 files changed, 28 insertions(+), 14 deletions(-)
diff --git a/include/tcg/tcg.h b/include/tcg/tcg.h
index f2da340bb9..8bcd60d0ed 100644
--- a/include/tcg/tcg.h
+++ b/include/tcg/tcg.h
@@ -319,6 +319,22 @@ typedef enum TCGType {
#endif
} TCGType;
+/**
+ * tcg_type_size
+ * @t: type
+ *
+ * Return the size of the type in bytes.
+ */
+static inline int tcg_type_size(TCGType t)
+{
+ unsigned i = t;
+ if (i >= TCG_TYPE_V64) {
+ tcg_debug_assert(i < TCG_TYPE_COUNT);
+ i -= TCG_TYPE_V64 - 1;
+ }
+ return 4 << i;
+}
+
/**
* get_alignment_bits
* @memop: MemOp value
diff --git a/tcg/tcg.c b/tcg/tcg.c
index 73d8b33b06..48b517a856 100644
--- a/tcg/tcg.c
+++ b/tcg/tcg.c
@@ -3110,22 +3110,22 @@ static void check_regs(TCGContext *s)
static void temp_allocate_frame(TCGContext *s, TCGTemp *ts)
{
- intptr_t off, size, align;
+ int size = tcg_type_size(ts->type);
+ int align;
+ intptr_t off;
switch (ts->type) {
case TCG_TYPE_I32:
- size = align = 4;
+ align = 4;
break;
case TCG_TYPE_I64:
case TCG_TYPE_V64:
- size = align = 8;
+ align = 8;
break;
case TCG_TYPE_V128:
- size = align = 16;
- break;
case TCG_TYPE_V256:
/* Note that we do not require aligned storage for V256. */
- size = 32, align = 16;
+ align = 16;
break;
default:
g_assert_not_reached();
@@ -3641,8 +3641,8 @@ static void tcg_reg_alloc_dup(TCGContext *s, const TCGOp *op)
TCGRegSet dup_out_regs, dup_in_regs;
TCGTemp *its, *ots;
TCGType itype, vtype;
- intptr_t endian_fixup;
unsigned vece;
+ int lowpart_ofs;
bool ok;
ots = arg_temp(op->args[0]);
@@ -3711,14 +3711,12 @@ static void tcg_reg_alloc_dup(TCGContext *s, const TCGOp *op)
/* fall through */
case TEMP_VAL_MEM:
-#if HOST_BIG_ENDIAN
- endian_fixup = itype == TCG_TYPE_I32 ? 4 : 8;
- endian_fixup -= 1 << vece;
-#else
- endian_fixup = 0;
-#endif
+ lowpart_ofs = 0;
+ if (HOST_BIG_ENDIAN) {
+ lowpart_ofs = tcg_type_size(itype) - (1 << vece);
+ }
if (tcg_out_dupm_vec(s, vtype, vece, ots->reg, its->mem_base->reg,
- its->mem_offset + endian_fixup)) {
+ its->mem_offset + lowpart_ofs)) {
goto done;
}
tcg_out_ld(s, itype, ots->reg, its->mem_base->reg, its->mem_offset);
--
2.34.1
^ permalink raw reply related [flat|nested] 43+ messages in thread* [PATCH v2 16/36] tcg: Introduce TCGCallReturnKind and TCGCallArgumentKind
2022-10-21 7:15 [PATCH v2 00/36] tcg: Support for Int128 with helpers Richard Henderson
` (14 preceding siblings ...)
2022-10-21 7:15 ` [PATCH v2 15/36] tcg: Introduce tcg_type_size Richard Henderson
@ 2022-10-21 7:15 ` Richard Henderson
2022-10-21 7:15 ` [PATCH v2 17/36] tcg: Replace TCG_TARGET_CALL_ALIGN_ARGS with TCG_TARGET_CALL_ARG_I64 Richard Henderson
` (19 subsequent siblings)
35 siblings, 0 replies; 43+ messages in thread
From: Richard Henderson @ 2022-10-21 7:15 UTC (permalink / raw)
To: qemu-devel
Prepare to replace a bunch of separate ifdefs with a
consistent way to describe the abi of a function call.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
tcg/tcg-internal.h | 15 +++++++++++++++
1 file changed, 15 insertions(+)
diff --git a/tcg/tcg-internal.h b/tcg/tcg-internal.h
index 2c06b5116a..f574743ff8 100644
--- a/tcg/tcg-internal.h
+++ b/tcg/tcg-internal.h
@@ -27,6 +27,21 @@
#define TCG_HIGHWATER 1024
+/*
+ * Describe the calling convention of a given argument type.
+ */
+typedef enum {
+ TCG_CALL_RET_NORMAL, /* by registers */
+} TCGCallReturnKind;
+
+typedef enum {
+ TCG_CALL_ARG_NORMAL, /* by registers (continuing onto stack) */
+ TCG_CALL_ARG_EVEN, /* like normal, but skipping odd slots */
+ TCG_CALL_ARG_EXTEND, /* for i32, as a sign/zero-extended i64 */
+ TCG_CALL_ARG_EXTEND_U, /* ... as a zero-extended i64 */
+ TCG_CALL_ARG_EXTEND_S, /* ... as a sign-extended i64 */
+} TCGCallArgumentKind;
+
typedef struct TCGHelperInfo {
void *func;
const char *name;
--
2.34.1
^ permalink raw reply related [flat|nested] 43+ messages in thread* [PATCH v2 17/36] tcg: Replace TCG_TARGET_CALL_ALIGN_ARGS with TCG_TARGET_CALL_ARG_I64
2022-10-21 7:15 [PATCH v2 00/36] tcg: Support for Int128 with helpers Richard Henderson
` (15 preceding siblings ...)
2022-10-21 7:15 ` [PATCH v2 16/36] tcg: Introduce TCGCallReturnKind and TCGCallArgumentKind Richard Henderson
@ 2022-10-21 7:15 ` Richard Henderson
2022-10-21 10:28 ` Philippe Mathieu-Daudé
2022-10-21 7:15 ` [PATCH v2 18/36] tcg: Replace TCG_TARGET_EXTEND_ARGS with TCG_TARGET_CALL_ARG_I32 Richard Henderson
` (18 subsequent siblings)
35 siblings, 1 reply; 43+ messages in thread
From: Richard Henderson @ 2022-10-21 7:15 UTC (permalink / raw)
To: qemu-devel
For 32-bit hosts when TCG_TARGET_CALL_ALIGN_ARGS was set, use
TCG_CALL_ARG_EVEN. For 64-bit hosts, TCG_TARGET_CALL_ALIGN_ARGS
was silently ignored, so always use TCG_CALL_ARG_NORMAL.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
tcg/aarch64/tcg-target.h | 2 +-
tcg/arm/tcg-target.h | 2 +-
tcg/i386/tcg-target.h | 1 +
tcg/loongarch64/tcg-target.h | 2 +-
tcg/mips/tcg-target.h | 3 ++-
tcg/riscv/tcg-target.h | 6 +++++-
tcg/s390x/tcg-target.h | 1 +
tcg/sparc64/tcg-target.h | 1 +
tcg/tci/tcg-target.h | 5 +++++
tcg/tcg.c | 6 ++++--
tcg/ppc/tcg-target.c.inc | 21 ++++++++-------------
11 files changed, 30 insertions(+), 20 deletions(-)
diff --git a/tcg/aarch64/tcg-target.h b/tcg/aarch64/tcg-target.h
index e145d50fef..d9dd777caa 100644
--- a/tcg/aarch64/tcg-target.h
+++ b/tcg/aarch64/tcg-target.h
@@ -51,8 +51,8 @@ typedef enum {
/* used for function call generation */
#define TCG_REG_CALL_STACK TCG_REG_SP
#define TCG_TARGET_STACK_ALIGN 16
-#define TCG_TARGET_CALL_ALIGN_ARGS 1
#define TCG_TARGET_CALL_STACK_OFFSET 0
+#define TCG_TARGET_CALL_ARG_I64 TCG_CALL_ARG_NORMAL
/* optional instructions */
#define TCG_TARGET_HAS_div_i32 1
diff --git a/tcg/arm/tcg-target.h b/tcg/arm/tcg-target.h
index 56c1ac4586..09dd0550aa 100644
--- a/tcg/arm/tcg-target.h
+++ b/tcg/arm/tcg-target.h
@@ -88,8 +88,8 @@ extern bool use_neon_instructions;
/* used for function call generation */
#define TCG_TARGET_STACK_ALIGN 8
-#define TCG_TARGET_CALL_ALIGN_ARGS 1
#define TCG_TARGET_CALL_STACK_OFFSET 0
+#define TCG_TARGET_CALL_ARG_I64 TCG_CALL_ARG_EVEN
/* optional instructions */
#define TCG_TARGET_HAS_ext8s_i32 1
diff --git a/tcg/i386/tcg-target.h b/tcg/i386/tcg-target.h
index 00fcbe297d..42628a2791 100644
--- a/tcg/i386/tcg-target.h
+++ b/tcg/i386/tcg-target.h
@@ -98,6 +98,7 @@ typedef enum {
#else
#define TCG_TARGET_CALL_STACK_OFFSET 0
#endif
+#define TCG_TARGET_CALL_ARG_I64 TCG_CALL_ARG_NORMAL
extern bool have_bmi1;
extern bool have_popcnt;
diff --git a/tcg/loongarch64/tcg-target.h b/tcg/loongarch64/tcg-target.h
index d58a6162f2..584cac3dc2 100644
--- a/tcg/loongarch64/tcg-target.h
+++ b/tcg/loongarch64/tcg-target.h
@@ -88,8 +88,8 @@ typedef enum {
/* used for function call generation */
#define TCG_REG_CALL_STACK TCG_REG_SP
#define TCG_TARGET_STACK_ALIGN 16
-#define TCG_TARGET_CALL_ALIGN_ARGS 1
#define TCG_TARGET_CALL_STACK_OFFSET 0
+#define TCG_TARGET_CALL_ARG_I64 TCG_CALL_ARG_NORMAL
/* optional instructions */
#define TCG_TARGET_HAS_movcond_i32 0
diff --git a/tcg/mips/tcg-target.h b/tcg/mips/tcg-target.h
index 7669213175..bb7312aed4 100644
--- a/tcg/mips/tcg-target.h
+++ b/tcg/mips/tcg-target.h
@@ -83,10 +83,11 @@ typedef enum {
#define TCG_TARGET_STACK_ALIGN 16
#if _MIPS_SIM == _ABIO32
# define TCG_TARGET_CALL_STACK_OFFSET 16
+# define TCG_TARGET_CALL_ARG_I64 TCG_CALL_ARG_EVEN
#else
# define TCG_TARGET_CALL_STACK_OFFSET 0
+# define TCG_TARGET_CALL_ARG_I64 TCG_CALL_ARG_NORMAL
#endif
-#define TCG_TARGET_CALL_ALIGN_ARGS 1
/* MOVN/MOVZ instructions detection */
#if (defined(__mips_isa_rev) && (__mips_isa_rev >= 1)) || \
diff --git a/tcg/riscv/tcg-target.h b/tcg/riscv/tcg-target.h
index 11c9b3e4f4..2ab4b8d04a 100644
--- a/tcg/riscv/tcg-target.h
+++ b/tcg/riscv/tcg-target.h
@@ -81,8 +81,12 @@ typedef enum {
/* used for function call generation */
#define TCG_REG_CALL_STACK TCG_REG_SP
#define TCG_TARGET_STACK_ALIGN 16
-#define TCG_TARGET_CALL_ALIGN_ARGS 1
#define TCG_TARGET_CALL_STACK_OFFSET 0
+#if TCG_TARGET_REG_BITS == 32
+#define TCG_TARGET_CALL_ARG_I64 TCG_CALL_ARG_EVEN
+#else
+#define TCG_TARGET_CALL_ARG_I64 TCG_CALL_ARG_NORMAL
+#endif
/* optional instructions */
#define TCG_TARGET_HAS_movcond_i32 0
diff --git a/tcg/s390x/tcg-target.h b/tcg/s390x/tcg-target.h
index 23e2063667..7e86791ff6 100644
--- a/tcg/s390x/tcg-target.h
+++ b/tcg/s390x/tcg-target.h
@@ -166,6 +166,7 @@ extern uint64_t s390_facilities[3];
/* used for function call generation */
#define TCG_TARGET_STACK_ALIGN 8
#define TCG_TARGET_CALL_STACK_OFFSET 160
+#define TCG_TARGET_CALL_ARG_I64 TCG_CALL_RET_NORMAL
#define TCG_TARGET_EXTEND_ARGS 1
#define TCG_TARGET_HAS_MEMORY_BSWAP 1
diff --git a/tcg/sparc64/tcg-target.h b/tcg/sparc64/tcg-target.h
index 8655acdbe5..44ac164b31 100644
--- a/tcg/sparc64/tcg-target.h
+++ b/tcg/sparc64/tcg-target.h
@@ -72,6 +72,7 @@ typedef enum {
#define TCG_TARGET_STACK_ALIGN 16
#define TCG_TARGET_CALL_STACK_OFFSET (128 + 6*8 + TCG_TARGET_STACK_BIAS)
#define TCG_TARGET_EXTEND_ARGS 1
+#define TCG_TARGET_CALL_ARG_I64 TCG_CALL_ARG_NORMAL
#if defined(__VIS__) && __VIS__ >= 0x300
#define use_vis3_instructions 1
diff --git a/tcg/tci/tcg-target.h b/tcg/tci/tcg-target.h
index ceb36c4f7a..e11c293906 100644
--- a/tcg/tci/tcg-target.h
+++ b/tcg/tci/tcg-target.h
@@ -158,6 +158,11 @@ typedef enum {
/* Used for function call generation. */
#define TCG_TARGET_CALL_STACK_OFFSET 0
#define TCG_TARGET_STACK_ALIGN 8
+#if TCG_TARGET_REG_BITS == 32
+# define TCG_TARGET_CALL_ARG_I64 TCG_CALL_ARG_EVEN
+#else
+# define TCG_TARGET_CALL_ARG_I64 TCG_CALL_ARG_NORMAL
+#endif
#define HAVE_TCG_QEMU_TB_EXEC
#define TCG_TARGET_NEED_POOL_LABELS
diff --git a/tcg/tcg.c b/tcg/tcg.c
index 48b517a856..9abff104bf 100644
--- a/tcg/tcg.c
+++ b/tcg/tcg.c
@@ -1541,9 +1541,11 @@ void tcg_gen_callN(void *func, TCGTemp *ret, int nargs, TCGTemp **args)
* for passing off to ffi_call.
*/
want_align = true;
-#elif defined(TCG_TARGET_CALL_ALIGN_ARGS)
+#else
/* Some targets want aligned 64 bit args */
- want_align = is_64bit;
+ if (is_64bit) {
+ want_align = TCG_TARGET_CALL_ARG_I64 == TCG_CALL_ARG_EVEN;
+ }
#endif
if (TCG_TARGET_REG_BITS < 64 && want_align && (real_args & 1)) {
diff --git a/tcg/ppc/tcg-target.c.inc b/tcg/ppc/tcg-target.c.inc
index e3dba47697..d1d6a40c6c 100644
--- a/tcg/ppc/tcg-target.c.inc
+++ b/tcg/ppc/tcg-target.c.inc
@@ -45,7 +45,9 @@
#endif
#ifdef _CALL_SYSV
-# define TCG_TARGET_CALL_ALIGN_ARGS 1
+# define TCG_TARGET_CALL_ARG_I64 TCG_CALL_ARG_EVEN
+#else
+# define TCG_TARGET_CALL_ARG_I64 TCG_CALL_ARG_NORMAL
#endif
/* For some memory operations, we need a scratch that isn't R0. For the AIX
@@ -2202,9 +2204,7 @@ static bool tcg_out_qemu_ld_slow_path(TCGContext *s, TCGLabelQemuLdst *lb)
lo = lb->addrlo_reg;
hi = lb->addrhi_reg;
if (TCG_TARGET_REG_BITS < TARGET_LONG_BITS) {
-#ifdef TCG_TARGET_CALL_ALIGN_ARGS
- arg |= 1;
-#endif
+ arg |= (TCG_TARGET_CALL_ARG_I64 == TCG_CALL_ARG_EVEN);
tcg_out_mov(s, TCG_TYPE_I32, arg++, hi);
tcg_out_mov(s, TCG_TYPE_I32, arg++, lo);
} else {
@@ -2250,9 +2250,7 @@ static bool tcg_out_qemu_st_slow_path(TCGContext *s, TCGLabelQemuLdst *lb)
lo = lb->addrlo_reg;
hi = lb->addrhi_reg;
if (TCG_TARGET_REG_BITS < TARGET_LONG_BITS) {
-#ifdef TCG_TARGET_CALL_ALIGN_ARGS
- arg |= 1;
-#endif
+ arg |= (TCG_TARGET_CALL_ARG_I64 == TCG_CALL_ARG_EVEN);
tcg_out_mov(s, TCG_TYPE_I32, arg++, hi);
tcg_out_mov(s, TCG_TYPE_I32, arg++, lo);
} else {
@@ -2266,9 +2264,7 @@ static bool tcg_out_qemu_st_slow_path(TCGContext *s, TCGLabelQemuLdst *lb)
if (TCG_TARGET_REG_BITS == 32) {
switch (s_bits) {
case MO_64:
-#ifdef TCG_TARGET_CALL_ALIGN_ARGS
- arg |= 1;
-#endif
+ arg |= (TCG_TARGET_CALL_ARG_I64 == TCG_CALL_ARG_EVEN);
tcg_out_mov(s, TCG_TYPE_I32, arg++, hi);
/* FALLTHRU */
case MO_32:
@@ -2324,9 +2320,8 @@ static bool tcg_out_fail_alignment(TCGContext *s, TCGLabelQemuLdst *l)
if (TCG_TARGET_REG_BITS < TARGET_LONG_BITS) {
TCGReg arg = TCG_REG_R4;
-#ifdef TCG_TARGET_CALL_ALIGN_ARGS
- arg |= 1;
-#endif
+
+ arg |= (TCG_TARGET_CALL_ARG_I64 == TCG_CALL_ARG_EVEN);
if (l->addrlo_reg != arg) {
tcg_out_mov(s, TCG_TYPE_I32, arg, l->addrhi_reg);
tcg_out_mov(s, TCG_TYPE_I32, arg + 1, l->addrlo_reg);
--
2.34.1
^ permalink raw reply related [flat|nested] 43+ messages in thread* Re: [PATCH v2 17/36] tcg: Replace TCG_TARGET_CALL_ALIGN_ARGS with TCG_TARGET_CALL_ARG_I64
2022-10-21 7:15 ` [PATCH v2 17/36] tcg: Replace TCG_TARGET_CALL_ALIGN_ARGS with TCG_TARGET_CALL_ARG_I64 Richard Henderson
@ 2022-10-21 10:28 ` Philippe Mathieu-Daudé
0 siblings, 0 replies; 43+ messages in thread
From: Philippe Mathieu-Daudé @ 2022-10-21 10:28 UTC (permalink / raw)
To: Richard Henderson, qemu-devel
On 21/10/22 09:15, Richard Henderson wrote:
> For 32-bit hosts when TCG_TARGET_CALL_ALIGN_ARGS was set, use
> TCG_CALL_ARG_EVEN. For 64-bit hosts, TCG_TARGET_CALL_ALIGN_ARGS
> was silently ignored, so always use TCG_CALL_ARG_NORMAL.
>
> Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
> ---
> tcg/aarch64/tcg-target.h | 2 +-
> tcg/arm/tcg-target.h | 2 +-
> tcg/i386/tcg-target.h | 1 +
> tcg/loongarch64/tcg-target.h | 2 +-
> tcg/mips/tcg-target.h | 3 ++-
> tcg/riscv/tcg-target.h | 6 +++++-
> tcg/s390x/tcg-target.h | 1 +
> tcg/sparc64/tcg-target.h | 1 +
> tcg/tci/tcg-target.h | 5 +++++
> tcg/tcg.c | 6 ++++--
> tcg/ppc/tcg-target.c.inc | 21 ++++++++-------------
> 11 files changed, 30 insertions(+), 20 deletions(-)
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
^ permalink raw reply [flat|nested] 43+ messages in thread
* [PATCH v2 18/36] tcg: Replace TCG_TARGET_EXTEND_ARGS with TCG_TARGET_CALL_ARG_I32
2022-10-21 7:15 [PATCH v2 00/36] tcg: Support for Int128 with helpers Richard Henderson
` (16 preceding siblings ...)
2022-10-21 7:15 ` [PATCH v2 17/36] tcg: Replace TCG_TARGET_CALL_ALIGN_ARGS with TCG_TARGET_CALL_ARG_I64 Richard Henderson
@ 2022-10-21 7:15 ` Richard Henderson
2022-10-21 10:33 ` Philippe Mathieu-Daudé
2022-10-21 7:15 ` [PATCH v2 19/36] tcg: Use TCG_CALL_ARG_EVEN for TCI special case Richard Henderson
` (17 subsequent siblings)
35 siblings, 1 reply; 43+ messages in thread
From: Richard Henderson @ 2022-10-21 7:15 UTC (permalink / raw)
To: qemu-devel
For 64-bit hosts that had TCG_TARGET_EXTEND_ARGS, set
TCG_TARGET_CALL_ARG_I32 to TCG_CALL_ARG_EXTEND.
Otherwise, use TCG_CALL_ARG_NORMAL.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
tcg/aarch64/tcg-target.h | 1 +
tcg/arm/tcg-target.h | 1 +
tcg/i386/tcg-target.h | 1 +
tcg/loongarch64/tcg-target.h | 1 +
tcg/mips/tcg-target.h | 1 +
tcg/riscv/tcg-target.h | 1 +
tcg/s390x/tcg-target.h | 2 +-
tcg/sparc64/tcg-target.h | 2 +-
tcg/tci/tcg-target.h | 1 +
tcg/tcg.c | 42 ++++++++++++++++++------------------
tcg/ppc/tcg-target.c.inc | 6 +++++-
11 files changed, 35 insertions(+), 24 deletions(-)
diff --git a/tcg/aarch64/tcg-target.h b/tcg/aarch64/tcg-target.h
index d9dd777caa..413a5410c5 100644
--- a/tcg/aarch64/tcg-target.h
+++ b/tcg/aarch64/tcg-target.h
@@ -52,6 +52,7 @@ typedef enum {
#define TCG_REG_CALL_STACK TCG_REG_SP
#define TCG_TARGET_STACK_ALIGN 16
#define TCG_TARGET_CALL_STACK_OFFSET 0
+#define TCG_TARGET_CALL_ARG_I32 TCG_CALL_ARG_NORMAL
#define TCG_TARGET_CALL_ARG_I64 TCG_CALL_ARG_NORMAL
/* optional instructions */
diff --git a/tcg/arm/tcg-target.h b/tcg/arm/tcg-target.h
index 09dd0550aa..b7843d2d54 100644
--- a/tcg/arm/tcg-target.h
+++ b/tcg/arm/tcg-target.h
@@ -89,6 +89,7 @@ extern bool use_neon_instructions;
/* used for function call generation */
#define TCG_TARGET_STACK_ALIGN 8
#define TCG_TARGET_CALL_STACK_OFFSET 0
+#define TCG_TARGET_CALL_ARG_I32 TCG_CALL_ARG_NORMAL
#define TCG_TARGET_CALL_ARG_I64 TCG_CALL_ARG_EVEN
/* optional instructions */
diff --git a/tcg/i386/tcg-target.h b/tcg/i386/tcg-target.h
index 42628a2791..7edb7f1d9a 100644
--- a/tcg/i386/tcg-target.h
+++ b/tcg/i386/tcg-target.h
@@ -98,6 +98,7 @@ typedef enum {
#else
#define TCG_TARGET_CALL_STACK_OFFSET 0
#endif
+#define TCG_TARGET_CALL_ARG_I32 TCG_CALL_ARG_NORMAL
#define TCG_TARGET_CALL_ARG_I64 TCG_CALL_ARG_NORMAL
extern bool have_bmi1;
diff --git a/tcg/loongarch64/tcg-target.h b/tcg/loongarch64/tcg-target.h
index 584cac3dc2..74fbb22dfd 100644
--- a/tcg/loongarch64/tcg-target.h
+++ b/tcg/loongarch64/tcg-target.h
@@ -89,6 +89,7 @@ typedef enum {
#define TCG_REG_CALL_STACK TCG_REG_SP
#define TCG_TARGET_STACK_ALIGN 16
#define TCG_TARGET_CALL_STACK_OFFSET 0
+#define TCG_TARGET_CALL_ARG_I32 TCG_CALL_ARG_NORMAL
#define TCG_TARGET_CALL_ARG_I64 TCG_CALL_ARG_NORMAL
/* optional instructions */
diff --git a/tcg/mips/tcg-target.h b/tcg/mips/tcg-target.h
index bb7312aed4..15721c3e42 100644
--- a/tcg/mips/tcg-target.h
+++ b/tcg/mips/tcg-target.h
@@ -88,6 +88,7 @@ typedef enum {
# define TCG_TARGET_CALL_STACK_OFFSET 0
# define TCG_TARGET_CALL_ARG_I64 TCG_CALL_ARG_NORMAL
#endif
+#define TCG_TARGET_CALL_ARG_I32 TCG_CALL_ARG_NORMAL
/* MOVN/MOVZ instructions detection */
#if (defined(__mips_isa_rev) && (__mips_isa_rev >= 1)) || \
diff --git a/tcg/riscv/tcg-target.h b/tcg/riscv/tcg-target.h
index 2ab4b8d04a..232537ccea 100644
--- a/tcg/riscv/tcg-target.h
+++ b/tcg/riscv/tcg-target.h
@@ -82,6 +82,7 @@ typedef enum {
#define TCG_REG_CALL_STACK TCG_REG_SP
#define TCG_TARGET_STACK_ALIGN 16
#define TCG_TARGET_CALL_STACK_OFFSET 0
+#define TCG_TARGET_CALL_ARG_I32 TCG_CALL_ARG_NORMAL
#if TCG_TARGET_REG_BITS == 32
#define TCG_TARGET_CALL_ARG_I64 TCG_CALL_ARG_EVEN
#else
diff --git a/tcg/s390x/tcg-target.h b/tcg/s390x/tcg-target.h
index 7e86791ff6..db5665c375 100644
--- a/tcg/s390x/tcg-target.h
+++ b/tcg/s390x/tcg-target.h
@@ -166,9 +166,9 @@ extern uint64_t s390_facilities[3];
/* used for function call generation */
#define TCG_TARGET_STACK_ALIGN 8
#define TCG_TARGET_CALL_STACK_OFFSET 160
+#define TCG_TARGET_CALL_ARG_I32 TCG_CALL_ARG_EXTEND
#define TCG_TARGET_CALL_ARG_I64 TCG_CALL_RET_NORMAL
-#define TCG_TARGET_EXTEND_ARGS 1
#define TCG_TARGET_HAS_MEMORY_BSWAP 1
#define TCG_TARGET_DEFAULT_MO (TCG_MO_ALL & ~TCG_MO_ST_LD)
diff --git a/tcg/sparc64/tcg-target.h b/tcg/sparc64/tcg-target.h
index 44ac164b31..0044ac8d78 100644
--- a/tcg/sparc64/tcg-target.h
+++ b/tcg/sparc64/tcg-target.h
@@ -71,7 +71,7 @@ typedef enum {
#define TCG_TARGET_STACK_BIAS 2047
#define TCG_TARGET_STACK_ALIGN 16
#define TCG_TARGET_CALL_STACK_OFFSET (128 + 6*8 + TCG_TARGET_STACK_BIAS)
-#define TCG_TARGET_EXTEND_ARGS 1
+#define TCG_TARGET_CALL_ARG_I32 TCG_CALL_ARG_EXTEND
#define TCG_TARGET_CALL_ARG_I64 TCG_CALL_ARG_NORMAL
#if defined(__VIS__) && __VIS__ >= 0x300
diff --git a/tcg/tci/tcg-target.h b/tcg/tci/tcg-target.h
index e11c293906..d6e0450ed8 100644
--- a/tcg/tci/tcg-target.h
+++ b/tcg/tci/tcg-target.h
@@ -158,6 +158,7 @@ typedef enum {
/* Used for function call generation. */
#define TCG_TARGET_CALL_STACK_OFFSET 0
#define TCG_TARGET_STACK_ALIGN 8
+#define TCG_TARGET_CALL_ARG_I32 TCG_CALL_ARG_NORMAL
#if TCG_TARGET_REG_BITS == 32
# define TCG_TARGET_CALL_ARG_I64 TCG_CALL_ARG_EVEN
#else
diff --git a/tcg/tcg.c b/tcg/tcg.c
index 9abff104bf..d4960c62b5 100644
--- a/tcg/tcg.c
+++ b/tcg/tcg.c
@@ -1493,24 +1493,24 @@ void tcg_gen_callN(void *func, TCGTemp *ret, int nargs, TCGTemp **args)
}
#endif
-#if defined(TCG_TARGET_EXTEND_ARGS) && TCG_TARGET_REG_BITS == 64
- for (i = 0; i < nargs; ++i) {
- int argtype = extract32(typemask, (i + 1) * 3, 3);
- bool is_32bit = (argtype & ~1) == dh_typecode_i32;
- bool is_signed = argtype & 1;
+ if (TCG_TARGET_CALL_ARG_I32 == TCG_CALL_ARG_EXTEND) {
+ for (i = 0; i < nargs; ++i) {
+ int argtype = extract32(typemask, (i + 1) * 3, 3);
+ bool is_32bit = (argtype & ~1) == dh_typecode_i32;
+ bool is_signed = argtype & 1;
- if (is_32bit) {
- TCGv_i64 temp = tcg_temp_new_i64();
- TCGv_i32 orig = temp_tcgv_i32(args[i]);
- if (is_signed) {
- tcg_gen_ext_i32_i64(temp, orig);
- } else {
- tcg_gen_extu_i32_i64(temp, orig);
+ if (is_32bit) {
+ TCGv_i64 temp = tcg_temp_new_i64();
+ TCGv_i32 orig = temp_tcgv_i32(args[i]);
+ if (is_signed) {
+ tcg_gen_ext_i32_i64(temp, orig);
+ } else {
+ tcg_gen_extu_i32_i64(temp, orig);
+ }
+ args[i] = tcgv_i64_temp(temp);
}
- args[i] = tcgv_i64_temp(temp);
}
}
-#endif /* TCG_TARGET_EXTEND_ARGS */
op = tcg_emit_op(INDEX_op_call);
@@ -1571,16 +1571,16 @@ void tcg_gen_callN(void *func, TCGTemp *ret, int nargs, TCGTemp **args)
tcg_debug_assert(TCGOP_CALLI(op) == real_args);
tcg_debug_assert(pi <= ARRAY_SIZE(op->args));
-#if defined(TCG_TARGET_EXTEND_ARGS) && TCG_TARGET_REG_BITS == 64
- for (i = 0; i < nargs; ++i) {
- int argtype = extract32(typemask, (i + 1) * 3, 3);
- bool is_32bit = (argtype & ~1) == dh_typecode_i32;
+ if (TCG_TARGET_CALL_ARG_I32 == TCG_CALL_ARG_EXTEND) {
+ for (i = 0; i < nargs; ++i) {
+ int argtype = extract32(typemask, (i + 1) * 3, 3);
+ bool is_32bit = (argtype & ~1) == dh_typecode_i32;
- if (is_32bit) {
- tcg_temp_free_internal(args[i]);
+ if (is_32bit) {
+ tcg_temp_free_internal(args[i]);
+ }
}
}
-#endif /* TCG_TARGET_EXTEND_ARGS */
}
static void tcg_reg_alloc_start(TCGContext *s)
diff --git a/tcg/ppc/tcg-target.c.inc b/tcg/ppc/tcg-target.c.inc
index d1d6a40c6c..500b75a03a 100644
--- a/tcg/ppc/tcg-target.c.inc
+++ b/tcg/ppc/tcg-target.c.inc
@@ -44,6 +44,11 @@
# endif
#endif
+#if TCG_TARGET_REG_BITS == 64
+# define TCG_TARGET_CALL_ARG_I32 TCG_CALL_ARG_EXTEND
+#else
+# define TCG_TARGET_CALL_ARG_I32 TCG_CALL_ARG_NORMAL
+#endif
#ifdef _CALL_SYSV
# define TCG_TARGET_CALL_ARG_I64 TCG_CALL_ARG_EVEN
#else
@@ -2520,7 +2525,6 @@ static void tcg_out_nop_fill(tcg_insn_unit *p, int count)
/* Parameters for function call generation, used in tcg.c. */
#define TCG_TARGET_STACK_ALIGN 16
-#define TCG_TARGET_EXTEND_ARGS 1
#ifdef _CALL_AIX
# define LINK_AREA_SIZE (6 * SZR)
--
2.34.1
^ permalink raw reply related [flat|nested] 43+ messages in thread* Re: [PATCH v2 18/36] tcg: Replace TCG_TARGET_EXTEND_ARGS with TCG_TARGET_CALL_ARG_I32
2022-10-21 7:15 ` [PATCH v2 18/36] tcg: Replace TCG_TARGET_EXTEND_ARGS with TCG_TARGET_CALL_ARG_I32 Richard Henderson
@ 2022-10-21 10:33 ` Philippe Mathieu-Daudé
0 siblings, 0 replies; 43+ messages in thread
From: Philippe Mathieu-Daudé @ 2022-10-21 10:33 UTC (permalink / raw)
To: Richard Henderson, qemu-devel
On 21/10/22 09:15, Richard Henderson wrote:
> For 64-bit hosts that had TCG_TARGET_EXTEND_ARGS, set
> TCG_TARGET_CALL_ARG_I32 to TCG_CALL_ARG_EXTEND.
> Otherwise, use TCG_CALL_ARG_NORMAL.
>
> Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
> ---
> tcg/aarch64/tcg-target.h | 1 +
> tcg/arm/tcg-target.h | 1 +
> tcg/i386/tcg-target.h | 1 +
> tcg/loongarch64/tcg-target.h | 1 +
> tcg/mips/tcg-target.h | 1 +
> tcg/riscv/tcg-target.h | 1 +
> tcg/s390x/tcg-target.h | 2 +-
> tcg/sparc64/tcg-target.h | 2 +-
> tcg/tci/tcg-target.h | 1 +
> tcg/tcg.c | 42 ++++++++++++++++++------------------
> tcg/ppc/tcg-target.c.inc | 6 +++++-
> 11 files changed, 35 insertions(+), 24 deletions(-)
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
^ permalink raw reply [flat|nested] 43+ messages in thread
* [PATCH v2 19/36] tcg: Use TCG_CALL_ARG_EVEN for TCI special case
2022-10-21 7:15 [PATCH v2 00/36] tcg: Support for Int128 with helpers Richard Henderson
` (17 preceding siblings ...)
2022-10-21 7:15 ` [PATCH v2 18/36] tcg: Replace TCG_TARGET_EXTEND_ARGS with TCG_TARGET_CALL_ARG_I32 Richard Henderson
@ 2022-10-21 7:15 ` Richard Henderson
2022-10-25 20:14 ` Ilya Leoshkevich
2022-10-21 7:15 ` [PATCH v2 20/36] tcg: Reorg function calls Richard Henderson
` (16 subsequent siblings)
35 siblings, 1 reply; 43+ messages in thread
From: Richard Henderson @ 2022-10-21 7:15 UTC (permalink / raw)
To: qemu-devel
Change 32-bit tci TCG_TARGET_CALL_ARG_I32 to TCG_CALL_ARG_EVEN, to
force 32-bit values to be aligned to 64-bit. With a small reorg
to the argument processing loop, this neatly replaces an ifdef for
CONFIG_TCG_INTERPRETER.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
tcg/tci/tcg-target.h | 3 ++-
tcg/tcg.c | 62 ++++++++++++++++++++++++++------------------
2 files changed, 39 insertions(+), 26 deletions(-)
diff --git a/tcg/tci/tcg-target.h b/tcg/tci/tcg-target.h
index d6e0450ed8..94ec541b4e 100644
--- a/tcg/tci/tcg-target.h
+++ b/tcg/tci/tcg-target.h
@@ -158,10 +158,11 @@ typedef enum {
/* Used for function call generation. */
#define TCG_TARGET_CALL_STACK_OFFSET 0
#define TCG_TARGET_STACK_ALIGN 8
-#define TCG_TARGET_CALL_ARG_I32 TCG_CALL_ARG_NORMAL
#if TCG_TARGET_REG_BITS == 32
+# define TCG_TARGET_CALL_ARG_I32 TCG_CALL_ARG_EVEN
# define TCG_TARGET_CALL_ARG_I64 TCG_CALL_ARG_EVEN
#else
+# define TCG_TARGET_CALL_ARG_I32 TCG_CALL_ARG_NORMAL
# define TCG_TARGET_CALL_ARG_I64 TCG_CALL_ARG_NORMAL
#endif
diff --git a/tcg/tcg.c b/tcg/tcg.c
index d4960c62b5..e0f5c6ea7b 100644
--- a/tcg/tcg.c
+++ b/tcg/tcg.c
@@ -1532,36 +1532,48 @@ void tcg_gen_callN(void *func, TCGTemp *ret, int nargs, TCGTemp **args)
real_args = 0;
for (i = 0; i < nargs; i++) {
int argtype = extract32(typemask, (i + 1) * 3, 3);
- bool is_64bit = (argtype & ~1) == dh_typecode_i64;
- bool want_align = false;
+ TCGCallArgumentKind kind;
+ TCGType type;
-#if defined(CONFIG_TCG_INTERPRETER)
- /*
- * Align all arguments, so that they land in predictable places
- * for passing off to ffi_call.
- */
- want_align = true;
-#else
- /* Some targets want aligned 64 bit args */
- if (is_64bit) {
- want_align = TCG_TARGET_CALL_ARG_I64 == TCG_CALL_ARG_EVEN;
- }
-#endif
-
- if (TCG_TARGET_REG_BITS < 64 && want_align && (real_args & 1)) {
- op->args[pi++] = TCG_CALL_DUMMY_ARG;
- real_args++;
+ switch (argtype) {
+ case dh_typecode_i32:
+ case dh_typecode_s32:
+ type = TCG_TYPE_I32;
+ kind = TCG_TARGET_CALL_ARG_I32;
+ break;
+ case dh_typecode_i64:
+ case dh_typecode_s64:
+ type = TCG_TYPE_I64;
+ kind = TCG_TARGET_CALL_ARG_I64;
+ break;
+ case dh_typecode_ptr:
+ type = TCG_TYPE_PTR;
+ kind = TCG_CALL_ARG_NORMAL;
+ break;
+ default:
+ g_assert_not_reached();
}
- if (TCG_TARGET_REG_BITS < 64 && is_64bit) {
+ switch (kind) {
+ case TCG_CALL_ARG_EVEN:
+ if (real_args & 1) {
+ op->args[pi++] = TCG_CALL_DUMMY_ARG;
+ real_args++;
+ }
+ /* fall through */
+ case TCG_CALL_ARG_NORMAL:
+ if (TCG_TARGET_REG_BITS == 32 && type == TCG_TYPE_I64) {
+ op->args[pi++] = temp_arg(args[i]);
+ op->args[pi++] = temp_arg(args[i] + 1);
+ real_args += 2;
+ break;
+ }
op->args[pi++] = temp_arg(args[i]);
- op->args[pi++] = temp_arg(args[i] + 1);
- real_args += 2;
- continue;
+ real_args++;
+ break;
+ default:
+ g_assert_not_reached();
}
-
- op->args[pi++] = temp_arg(args[i]);
- real_args++;
}
op->args[pi++] = (uintptr_t)func;
op->args[pi++] = (uintptr_t)info;
--
2.34.1
^ permalink raw reply related [flat|nested] 43+ messages in thread* Re: [PATCH v2 19/36] tcg: Use TCG_CALL_ARG_EVEN for TCI special case
2022-10-21 7:15 ` [PATCH v2 19/36] tcg: Use TCG_CALL_ARG_EVEN for TCI special case Richard Henderson
@ 2022-10-25 20:14 ` Ilya Leoshkevich
0 siblings, 0 replies; 43+ messages in thread
From: Ilya Leoshkevich @ 2022-10-25 20:14 UTC (permalink / raw)
To: Richard Henderson, qemu-devel
On Fri, Oct 21, 2022 at 05:15:32PM +1000, Richard Henderson wrote:
> Change 32-bit tci TCG_TARGET_CALL_ARG_I32 to TCG_CALL_ARG_EVEN, to
> force 32-bit values to be aligned to 64-bit. With a small reorg
> to the argument processing loop, this neatly replaces an ifdef for
> CONFIG_TCG_INTERPRETER.
>
> Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
> ---
> tcg/tci/tcg-target.h | 3 ++-
> tcg/tcg.c | 62 ++++++++++++++++++++++++++------------------
> 2 files changed, 39 insertions(+), 26 deletions(-)
Hi,
This patch breaks s390x-on-s390x emulation:
s390x$ ./configure --target-list=s390x-linux-user
s390x$ make -j$(nproc)
s390x$ make check-tcg
...
TEST convd on s390x
ERROR:../tcg/tcg.c:1575:tcg_gen_callN: code should not be reached
With the next patch ("tcg: Reorg function calls") the failure is
slightly different:
TEST convd on s390x
timeout: the monitored command dumped core
s390x-on-x86_64 works though.
Best regards,
Ilya
^ permalink raw reply [flat|nested] 43+ messages in thread
* [PATCH v2 20/36] tcg: Reorg function calls
2022-10-21 7:15 [PATCH v2 00/36] tcg: Support for Int128 with helpers Richard Henderson
` (18 preceding siblings ...)
2022-10-21 7:15 ` [PATCH v2 19/36] tcg: Use TCG_CALL_ARG_EVEN for TCI special case Richard Henderson
@ 2022-10-21 7:15 ` Richard Henderson
2022-10-21 7:15 ` [PATCH v2 21/36] tcg: Move ffi_cif pointer into TCGHelperInfo Richard Henderson
` (15 subsequent siblings)
35 siblings, 0 replies; 43+ messages in thread
From: Richard Henderson @ 2022-10-21 7:15 UTC (permalink / raw)
To: qemu-devel
Pre-compute the function call layout for each helper at startup.
Drop TCG_CALL_DUMMY_ARG, as we no longer need to leave gaps
in the op->args[] array.
For tcg_gen_callN, loop over the arguments once. Allocate the TCGOp
for the call early but delay emitting it, collecting arguments first.
This allows the argument processing loop to emit code for extensions
and have them sequenced before the call. Free the temporaries for
extensions immediately.
For tcg_reg_alloc_call, loop over the arguments in reverse order,
which allows stack slots to be filled first naturally.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
include/tcg/tcg.h | 3 -
tcg/tcg-internal.h | 17 +-
tcg/tcg.c | 591 +++++++++++++++++++++++++++------------------
3 files changed, 377 insertions(+), 234 deletions(-)
diff --git a/include/tcg/tcg.h b/include/tcg/tcg.h
index 8bcd60d0ed..8d0626c797 100644
--- a/include/tcg/tcg.h
+++ b/include/tcg/tcg.h
@@ -425,9 +425,6 @@ typedef TCGv_ptr TCGv_env;
#define TCG_CALL_NO_RWG_SE (TCG_CALL_NO_RWG | TCG_CALL_NO_SE)
#define TCG_CALL_NO_WG_SE (TCG_CALL_NO_WG | TCG_CALL_NO_SE)
-/* Used to align parameters. See the comment before tcgv_i32_temp. */
-#define TCG_CALL_DUMMY_ARG ((TCGArg)0)
-
/*
* Flags for the bswap opcodes.
* If IZ, the input is zero-extended, otherwise unknown.
diff --git a/tcg/tcg-internal.h b/tcg/tcg-internal.h
index f574743ff8..097fef2325 100644
--- a/tcg/tcg-internal.h
+++ b/tcg/tcg-internal.h
@@ -42,11 +42,24 @@ typedef enum {
TCG_CALL_ARG_EXTEND_S, /* ... as a sign-extended i64 */
} TCGCallArgumentKind;
+typedef struct TCGCallArgumentLoc {
+ TCGCallArgumentKind kind : 8;
+ unsigned reg_slot : 8;
+ unsigned stk_slot : 8;
+ unsigned reg_n : 2;
+ unsigned arg_idx : 4;
+ unsigned tmp_subindex : 1;
+} TCGCallArgumentLoc;
+
typedef struct TCGHelperInfo {
void *func;
const char *name;
- unsigned flags;
- unsigned typemask;
+ unsigned typemask : 32;
+ unsigned flags : 8;
+ unsigned nr_in : 8;
+ unsigned nr_out : 8;
+ TCGCallReturnKind out_kind : 8;
+ TCGCallArgumentLoc in[MAX_OPC_PARAM_IARGS * MAX_OPC_PARAM_PER_ARG];
} TCGHelperInfo;
extern TCGContext tcg_init_ctx;
diff --git a/tcg/tcg.c b/tcg/tcg.c
index e0f5c6ea7b..713e692621 100644
--- a/tcg/tcg.c
+++ b/tcg/tcg.c
@@ -546,7 +546,7 @@ void tcg_pool_reset(TCGContext *s)
#include "exec/helper-proto.h"
-static const TCGHelperInfo all_helpers[] = {
+static TCGHelperInfo all_helpers[] = {
#include "exec/helper-tcg.h"
};
static GHashTable *helper_table;
@@ -564,6 +564,162 @@ static ffi_type * const typecode_to_ffi[8] = {
};
#endif
+typedef struct TCGCumulativeArgs {
+ int arg_idx; /* tcg_gen_callN args[] */
+ int op_arg_idx; /* TCGOp op->args[] */
+ int info_in_idx; /* TCGHelperInfo in[] */
+ int reg_slot; /* tcg_target_call_iarg_regs[] */
+ int stk_slot; /* TCG_TARGET_CALL_STACK_OFFSET + i*word */
+ int max_reg_slot;
+ int max_stk_slot;
+} TCGCumulativeArgs;
+
+static void layout_arg_1(TCGCumulativeArgs *cum, TCGHelperInfo *info,
+ TCGCallArgumentKind kind)
+{
+ TCGCallArgumentLoc *loc = &info->in[cum->info_in_idx];
+
+ *loc = (TCGCallArgumentLoc){
+ .kind = kind,
+ .arg_idx = cum->arg_idx,
+ };
+
+ if (cum->reg_slot < cum->max_reg_slot) {
+ loc->reg_slot = cum->reg_slot++;
+ loc->reg_n = 1;
+ } else {
+ loc->stk_slot = cum->stk_slot++;
+ }
+
+ cum->info_in_idx++;
+ cum->op_arg_idx++;
+}
+
+static void layout_arg_normal_2(TCGCumulativeArgs *cum, TCGHelperInfo *info)
+{
+ TCGCallArgumentLoc *loc = &info->in[cum->info_in_idx];
+
+ /* Layout the pair using the same arg_idx... */
+ layout_arg_1(cum, info, TCG_CALL_ARG_NORMAL);
+ layout_arg_1(cum, info, TCG_CALL_ARG_NORMAL);
+
+ /* ... then adjust the second loc to the second subindex. */
+ loc[1].tmp_subindex = 1;
+}
+
+static void init_call_layout(TCGHelperInfo *info)
+{
+ unsigned typemask = info->typemask;
+ unsigned typecode;
+ TCGCumulativeArgs cum = {
+ .max_reg_slot = ARRAY_SIZE(tcg_target_call_iarg_regs),
+ .max_stk_slot = TCG_STATIC_CALL_ARGS_SIZE / sizeof(tcg_target_long),
+ };
+
+ /*
+ * Parse and place any function return value.
+ */
+ typecode = typemask & 7;
+ switch (typecode) {
+ case dh_typecode_void:
+ info->nr_out = 0;
+ break;
+ case dh_typecode_i32:
+ case dh_typecode_s32:
+ case dh_typecode_ptr:
+ info->nr_out = 1;
+ info->out_kind = TCG_CALL_RET_NORMAL;
+ break;
+ case dh_typecode_i64:
+ case dh_typecode_s64:
+ info->nr_out = 64 / TCG_TARGET_REG_BITS;
+ info->out_kind = TCG_CALL_RET_NORMAL;
+ break;
+ default:
+ g_assert_not_reached();
+ }
+ assert(info->nr_out <= ARRAY_SIZE(tcg_target_call_oarg_regs));
+
+ /*
+ * The final two op->arg[] indexes are used for func & info.
+ * Account for that now to simplify the test below.
+ */
+ cum.op_arg_idx = info->nr_out + 2;
+
+ /*
+ * Parse and place function arguments.
+ */
+ for (typemask >>= 3; typemask; typemask >>= 3, cum.arg_idx++) {
+ TCGCallArgumentKind kind;
+ TCGType type;
+
+ typecode = typemask & 7;
+ switch (typecode) {
+ case dh_typecode_i32:
+ case dh_typecode_s32:
+ type = TCG_TYPE_I32;
+ kind = TCG_TARGET_CALL_ARG_I32;
+ break;
+ case dh_typecode_i64:
+ case dh_typecode_s64:
+ type = TCG_TYPE_I64;
+ kind = TCG_TARGET_CALL_ARG_I64;
+ break;
+ case dh_typecode_ptr:
+ type = TCG_TYPE_PTR;
+ kind = TCG_CALL_ARG_NORMAL;
+ break;
+ default:
+ g_assert_not_reached();
+ }
+
+ switch (kind) {
+ case TCG_CALL_ARG_EVEN:
+ tcg_debug_assert(cum.max_reg_slot % 2 == 0);
+ if (cum.reg_slot < cum.max_reg_slot) {
+ cum.reg_slot += cum.reg_slot & 1;
+ } else {
+ cum.stk_slot += cum.stk_slot & 1;
+ }
+ /* fall through */
+ case TCG_CALL_ARG_NORMAL:
+ switch (type) {
+ case TCG_TYPE_I32:
+ layout_arg_1(&cum, info, TCG_CALL_ARG_NORMAL);
+ break;
+ case TCG_TYPE_I64:
+ if (TCG_TARGET_REG_BITS == 32) {
+ layout_arg_normal_2(&cum, info);
+ } else {
+ layout_arg_1(&cum, info, TCG_CALL_ARG_NORMAL);
+ }
+ break;
+ default:
+ g_assert_not_reached();
+ }
+ break;
+ case TCG_CALL_ARG_EXTEND:
+ kind = TCG_CALL_ARG_EXTEND_U + (typecode & 1);
+ /* fall through */
+ case TCG_CALL_ARG_EXTEND_U:
+ case TCG_CALL_ARG_EXTEND_S:
+ assert(type == TCG_TYPE_I32);
+ layout_arg_1(&cum, info, kind);
+ break;
+ default:
+ g_assert_not_reached();
+ }
+ }
+ info->nr_in = cum.info_in_idx;
+
+ /* Validate that we didn't overrun the input array. */
+ assert(cum.info_in_idx <= ARRAY_SIZE(info->in));
+ /* Validate that we didn't overrun the output array. */
+ assert(cum.op_arg_idx <= MAX_OPC_PARAM);
+ /* Validate the backend has preallocated enough space on stack. */
+ assert(cum.stk_slot <= cum.max_stk_slot);
+}
+
static int indirect_reg_alloc_order[ARRAY_SIZE(tcg_target_reg_alloc_order)];
static void process_op_defs(TCGContext *s);
static TCGTemp *tcg_global_reg_new_internal(TCGContext *s, TCGType type,
@@ -603,6 +759,7 @@ static void tcg_context_init(unsigned max_cpus)
helper_table = g_hash_table_new(NULL, NULL);
for (i = 0; i < ARRAY_SIZE(all_helpers); ++i) {
+ init_call_layout(&all_helpers[i]);
g_hash_table_insert(helper_table, (gpointer)all_helpers[i].func,
(gpointer)&all_helpers[i]);
}
@@ -1473,18 +1630,15 @@ bool tcg_op_supported(TCGOpcode op)
}
}
-/* Note: we convert the 64 bit args to 32 bit and do some alignment
- and endian swap. Maybe it would be better to do the alignment
- and endian swap in tcg_reg_alloc_call(). */
+static TCGOp *tcg_op_alloc(TCGOpcode opc);
+
void tcg_gen_callN(void *func, TCGTemp *ret, int nargs, TCGTemp **args)
{
- int i, real_args, nb_rets, pi;
- unsigned typemask;
+ TCGOp *op = tcg_op_alloc(INDEX_op_call);
const TCGHelperInfo *info;
- TCGOp *op;
+ int i, n, pi = 0;
info = g_hash_table_lookup(helper_table, (gpointer)func);
- typemask = info->typemask;
#ifdef CONFIG_PLUGIN
/* detect non-plugin helpers */
@@ -1493,106 +1647,59 @@ void tcg_gen_callN(void *func, TCGTemp *ret, int nargs, TCGTemp **args)
}
#endif
- if (TCG_TARGET_CALL_ARG_I32 == TCG_CALL_ARG_EXTEND) {
- for (i = 0; i < nargs; ++i) {
- int argtype = extract32(typemask, (i + 1) * 3, 3);
- bool is_32bit = (argtype & ~1) == dh_typecode_i32;
- bool is_signed = argtype & 1;
+ TCGOP_CALLO(op) = n = info->nr_out;
+ switch (n) {
+ case 0:
+ tcg_debug_assert(ret == NULL);
+ break;
+ case 1:
+ tcg_debug_assert(ret != NULL);
+ op->args[pi++] = temp_arg(ret);
+ break;
+ case 2:
+ tcg_debug_assert(ret != NULL);
+ tcg_debug_assert(ret->temp_subindex == 0);
+ op->args[pi++] = temp_arg(ret);
+ op->args[pi++] = temp_arg(ret + 1);
+ break;
+ default:
+ g_assert_not_reached();
+ }
- if (is_32bit) {
+ TCGOP_CALLI(op) = n = info->nr_in;
+ for (i = 0; i < n; i++) {
+ const TCGCallArgumentLoc *loc = &info->in[i];
+ TCGTemp *ts = args[loc->arg_idx] + loc->tmp_subindex;
+
+ switch (loc->kind) {
+ case TCG_CALL_ARG_NORMAL:
+ op->args[pi++] = temp_arg(ts);
+ break;
+
+ case TCG_CALL_ARG_EXTEND_U:
+ case TCG_CALL_ARG_EXTEND_S:
+ {
TCGv_i64 temp = tcg_temp_new_i64();
- TCGv_i32 orig = temp_tcgv_i32(args[i]);
- if (is_signed) {
+ TCGv_i32 orig = temp_tcgv_i32(ts);
+
+ if (loc->kind == TCG_CALL_ARG_EXTEND_S) {
tcg_gen_ext_i32_i64(temp, orig);
} else {
tcg_gen_extu_i32_i64(temp, orig);
}
- args[i] = tcgv_i64_temp(temp);
+ op->args[pi++] = tcgv_i64_arg(temp);
+ tcg_temp_free_i64(temp);
}
- }
- }
-
- op = tcg_emit_op(INDEX_op_call);
-
- pi = 0;
- if (ret != NULL) {
- if (TCG_TARGET_REG_BITS < 64 && (typemask & 6) == dh_typecode_i64) {
- op->args[pi++] = temp_arg(ret);
- op->args[pi++] = temp_arg(ret + 1);
- nb_rets = 2;
- } else {
- op->args[pi++] = temp_arg(ret);
- nb_rets = 1;
- }
- } else {
- nb_rets = 0;
- }
- TCGOP_CALLO(op) = nb_rets;
-
- real_args = 0;
- for (i = 0; i < nargs; i++) {
- int argtype = extract32(typemask, (i + 1) * 3, 3);
- TCGCallArgumentKind kind;
- TCGType type;
-
- switch (argtype) {
- case dh_typecode_i32:
- case dh_typecode_s32:
- type = TCG_TYPE_I32;
- kind = TCG_TARGET_CALL_ARG_I32;
break;
- case dh_typecode_i64:
- case dh_typecode_s64:
- type = TCG_TYPE_I64;
- kind = TCG_TARGET_CALL_ARG_I64;
- break;
- case dh_typecode_ptr:
- type = TCG_TYPE_PTR;
- kind = TCG_CALL_ARG_NORMAL;
- break;
- default:
- g_assert_not_reached();
- }
- switch (kind) {
- case TCG_CALL_ARG_EVEN:
- if (real_args & 1) {
- op->args[pi++] = TCG_CALL_DUMMY_ARG;
- real_args++;
- }
- /* fall through */
- case TCG_CALL_ARG_NORMAL:
- if (TCG_TARGET_REG_BITS == 32 && type == TCG_TYPE_I64) {
- op->args[pi++] = temp_arg(args[i]);
- op->args[pi++] = temp_arg(args[i] + 1);
- real_args += 2;
- break;
- }
- op->args[pi++] = temp_arg(args[i]);
- real_args++;
- break;
default:
g_assert_not_reached();
}
}
op->args[pi++] = (uintptr_t)func;
op->args[pi++] = (uintptr_t)info;
- TCGOP_CALLI(op) = real_args;
- /* Make sure the fields didn't overflow. */
- tcg_debug_assert(TCGOP_CALLI(op) == real_args);
- tcg_debug_assert(pi <= ARRAY_SIZE(op->args));
-
- if (TCG_TARGET_CALL_ARG_I32 == TCG_CALL_ARG_EXTEND) {
- for (i = 0; i < nargs; ++i) {
- int argtype = extract32(typemask, (i + 1) * 3, 3);
- bool is_32bit = (argtype & ~1) == dh_typecode_i32;
-
- if (is_32bit) {
- tcg_temp_free_internal(args[i]);
- }
- }
- }
+ QTAILQ_INSERT_TAIL(&tcg_ctx->ops, op, link);
}
static void tcg_reg_alloc_start(TCGContext *s)
@@ -1807,10 +1914,7 @@ static void tcg_dump_ops(TCGContext *s, FILE *f, bool have_prefs)
}
for (i = 0; i < nb_iargs; i++) {
TCGArg arg = op->args[nb_oargs + i];
- const char *t = "<dummy>";
- if (arg != TCG_CALL_DUMMY_ARG) {
- t = tcg_get_arg_str(s, buf, sizeof(buf), arg);
- }
+ const char *t = tcg_get_arg_str(s, buf, sizeof(buf), arg);
col += ne_fprintf(f, ",%s", t);
}
} else {
@@ -2576,12 +2680,11 @@ static void liveness_pass_1(TCGContext *s)
switch (opc) {
case INDEX_op_call:
{
- int call_flags;
- int nb_call_regs;
+ const TCGHelperInfo *info = tcg_call_info(op);
+ int call_flags = tcg_call_flags(op);
nb_oargs = TCGOP_CALLO(op);
nb_iargs = TCGOP_CALLI(op);
- call_flags = tcg_call_flags(op);
/* pure functions can be removed if their result is unused */
if (call_flags & TCG_CALL_NO_SIDE_EFFECTS) {
@@ -2621,7 +2724,7 @@ static void liveness_pass_1(TCGContext *s)
/* Record arguments that die in this helper. */
for (i = nb_oargs; i < nb_iargs + nb_oargs; i++) {
ts = arg_temp(op->args[i]);
- if (ts && ts->state & TS_DEAD) {
+ if (ts->state & TS_DEAD) {
arg_life |= DEAD_ARG << i;
}
}
@@ -2629,31 +2732,59 @@ static void liveness_pass_1(TCGContext *s)
/* For all live registers, remove call-clobbered prefs. */
la_cross_call(s, nb_temps);
- nb_call_regs = ARRAY_SIZE(tcg_target_call_iarg_regs);
+ /*
+ * Input arguments are live for preceding opcodes.
+ *
+ * For those arguments that die, and will be allocated in
+ * registers, clear the register set for that arg, to be
+ * filled in below. For args that will be on the stack,
+ * reset to any available reg. Process arguments in reverse
+ * order so that if a temp is used more than once, the stack
+ * reset to max happens before the register reset to 0.
+ */
+ for (i = nb_iargs - 1; i >= 0; i--) {
+ const TCGCallArgumentLoc *loc = &info->in[i];
+ ts = arg_temp(op->args[nb_oargs + i]);
- /* Input arguments are live for preceding opcodes. */
- for (i = 0; i < nb_iargs; i++) {
- ts = arg_temp(op->args[i + nb_oargs]);
- if (ts && ts->state & TS_DEAD) {
- /* For those arguments that die, and will be allocated
- * in registers, clear the register set for that arg,
- * to be filled in below. For args that will be on
- * the stack, reset to any available reg.
- */
- *la_temp_pref(ts)
- = (i < nb_call_regs ? 0 :
- tcg_target_available_regs[ts->type]);
+ if (ts->state & TS_DEAD) {
+ switch (loc->kind) {
+ case TCG_CALL_ARG_NORMAL:
+ case TCG_CALL_ARG_EXTEND_U:
+ case TCG_CALL_ARG_EXTEND_S:
+ if (loc->reg_n) {
+ *la_temp_pref(ts) = 0;
+ break;
+ }
+ /* fall through */
+ default:
+ *la_temp_pref(ts) =
+ tcg_target_available_regs[ts->type];
+ break;
+ }
ts->state &= ~TS_DEAD;
}
}
- /* For each input argument, add its input register to prefs.
- If a temp is used once, this produces a single set bit. */
- for (i = 0; i < MIN(nb_call_regs, nb_iargs); i++) {
- ts = arg_temp(op->args[i + nb_oargs]);
- if (ts) {
- tcg_regset_set_reg(*la_temp_pref(ts),
- tcg_target_call_iarg_regs[i]);
+ /*
+ * For each input argument, add its input register to prefs.
+ * If a temp is used once, this produces a single set bit;
+ * if a temp is used multiple times, this produces a set.
+ */
+ for (i = 0; i < nb_iargs; i++) {
+ const TCGCallArgumentLoc *loc = &info->in[i];
+ ts = arg_temp(op->args[nb_oargs + i]);
+
+ switch (loc->kind) {
+ case TCG_CALL_ARG_NORMAL:
+ case TCG_CALL_ARG_EXTEND_U:
+ case TCG_CALL_ARG_EXTEND_S:
+ if (loc->reg_n) {
+ tcg_regset_set_reg(*la_temp_pref(ts),
+ tcg_target_call_iarg_regs[loc->reg_slot]);
+ }
+ break;
+ default:
+ break;
}
}
}
@@ -2922,21 +3053,19 @@ static bool liveness_pass_2(TCGContext *s)
/* Make sure that input arguments are available. */
for (i = nb_oargs; i < nb_iargs + nb_oargs; i++) {
arg_ts = arg_temp(op->args[i]);
- if (arg_ts) {
- dir_ts = arg_ts->state_ptr;
- if (dir_ts && arg_ts->state == TS_DEAD) {
- TCGOpcode lopc = (arg_ts->type == TCG_TYPE_I32
- ? INDEX_op_ld_i32
- : INDEX_op_ld_i64);
- TCGOp *lop = tcg_op_insert_before(s, op, lopc);
+ dir_ts = arg_ts->state_ptr;
+ if (dir_ts && arg_ts->state == TS_DEAD) {
+ TCGOpcode lopc = (arg_ts->type == TCG_TYPE_I32
+ ? INDEX_op_ld_i32
+ : INDEX_op_ld_i64);
+ TCGOp *lop = tcg_op_insert_before(s, op, lopc);
- lop->args[0] = temp_arg(dir_ts);
- lop->args[1] = temp_arg(arg_ts->mem_base);
- lop->args[2] = arg_ts->mem_offset;
+ lop->args[0] = temp_arg(dir_ts);
+ lop->args[1] = temp_arg(arg_ts->mem_base);
+ lop->args[2] = arg_ts->mem_offset;
- /* Loaded, but synced with memory. */
- arg_ts->state = TS_MEM;
- }
+ /* Loaded, but synced with memory. */
+ arg_ts->state = TS_MEM;
}
}
@@ -4158,106 +4287,100 @@ static bool tcg_reg_alloc_dup2(TCGContext *s, const TCGOp *op)
return true;
}
+static void load_arg_normal_1(TCGContext *s, const TCGCallArgumentLoc *loc,
+ TCGTemp *ts, TCGRegSet *allocated_regs)
+{
+ TCGReg reg;
+
+ /*
+ * If the destination is on the stack, load up the temp and store.
+ * If there are many call-saved registers, the temp might live to
+ * see another use; otherwise it'll be discarded.
+ */
+ if (!loc->reg_n) {
+ temp_load(s, ts, tcg_target_available_regs[ts->type],
+ *allocated_regs, 0);
+ tcg_out_st(s, ts->type, ts->reg, TCG_REG_CALL_STACK,
+ TCG_TARGET_CALL_STACK_OFFSET +
+ loc->stk_slot * sizeof(tcg_target_long));
+ return;
+ }
+
+ reg = tcg_target_call_iarg_regs[loc->reg_slot];
+
+ if (ts->val_type == TEMP_VAL_REG) {
+ if (ts->reg != reg) {
+ tcg_reg_free(s, reg, *allocated_regs);
+ if (!tcg_out_mov(s, ts->type, reg, ts->reg)) {
+ /*
+ * Cross register class move not supported. Sync the
+ * temp back to its slot and load from there.
+ */
+ temp_sync(s, ts, *allocated_regs, 0, 0);
+ tcg_out_ld(s, ts->type, reg,
+ ts->mem_base->reg, ts->mem_offset);
+ }
+ }
+ } else {
+ TCGRegSet arg_set = 0;
+
+ tcg_reg_free(s, reg, *allocated_regs);
+ tcg_regset_set_reg(arg_set, reg);
+ temp_load(s, ts, arg_set, *allocated_regs, 0);
+ }
+
+ tcg_regset_set_reg(*allocated_regs, reg);
+}
+
static void tcg_reg_alloc_call(TCGContext *s, TCGOp *op)
{
const int nb_oargs = TCGOP_CALLO(op);
const int nb_iargs = TCGOP_CALLI(op);
const TCGLifeData arg_life = op->life;
- const TCGHelperInfo *info;
- int flags, nb_regs, i;
- TCGReg reg;
- TCGArg arg;
- TCGTemp *ts;
- intptr_t stack_offset;
- size_t call_stack_size;
- tcg_insn_unit *func_addr;
- int allocate_args;
- TCGRegSet allocated_regs;
+ const TCGHelperInfo *info = tcg_call_info(op);
+ TCGRegSet allocated_regs = s->reserved_regs;
+ int i;
- func_addr = tcg_call_func(op);
- info = tcg_call_info(op);
- flags = info->flags;
+ /*
+ * Move inputs into place in reverse order,
+ * so that we place stacked arguments first.
+ */
+ for (i = nb_iargs - 1; i >= 0; --i) {
+ const TCGCallArgumentLoc *loc = &info->in[i];
+ TCGTemp *ts = arg_temp(op->args[nb_oargs + i]);
- nb_regs = ARRAY_SIZE(tcg_target_call_iarg_regs);
- if (nb_regs > nb_iargs) {
- nb_regs = nb_iargs;
- }
-
- /* assign stack slots first */
- call_stack_size = (nb_iargs - nb_regs) * sizeof(tcg_target_long);
- call_stack_size = (call_stack_size + TCG_TARGET_STACK_ALIGN - 1) &
- ~(TCG_TARGET_STACK_ALIGN - 1);
- allocate_args = (call_stack_size > TCG_STATIC_CALL_ARGS_SIZE);
- if (allocate_args) {
- /* XXX: if more than TCG_STATIC_CALL_ARGS_SIZE is needed,
- preallocate call stack */
- tcg_abort();
- }
-
- stack_offset = TCG_TARGET_CALL_STACK_OFFSET;
- for (i = nb_regs; i < nb_iargs; i++) {
- arg = op->args[nb_oargs + i];
- if (arg != TCG_CALL_DUMMY_ARG) {
- ts = arg_temp(arg);
- temp_load(s, ts, tcg_target_available_regs[ts->type],
- s->reserved_regs, 0);
- tcg_out_st(s, ts->type, ts->reg, TCG_REG_CALL_STACK, stack_offset);
- }
- stack_offset += sizeof(tcg_target_long);
- }
-
- /* assign input registers */
- allocated_regs = s->reserved_regs;
- for (i = 0; i < nb_regs; i++) {
- arg = op->args[nb_oargs + i];
- if (arg != TCG_CALL_DUMMY_ARG) {
- ts = arg_temp(arg);
- reg = tcg_target_call_iarg_regs[i];
-
- if (ts->val_type == TEMP_VAL_REG) {
- if (ts->reg != reg) {
- tcg_reg_free(s, reg, allocated_regs);
- if (!tcg_out_mov(s, ts->type, reg, ts->reg)) {
- /*
- * Cross register class move not supported. Sync the
- * temp back to its slot and load from there.
- */
- temp_sync(s, ts, allocated_regs, 0, 0);
- tcg_out_ld(s, ts->type, reg,
- ts->mem_base->reg, ts->mem_offset);
- }
- }
- } else {
- TCGRegSet arg_set = 0;
-
- tcg_reg_free(s, reg, allocated_regs);
- tcg_regset_set_reg(arg_set, reg);
- temp_load(s, ts, arg_set, allocated_regs, 0);
- }
-
- tcg_regset_set_reg(allocated_regs, reg);
+ switch (loc->kind) {
+ case TCG_CALL_ARG_NORMAL:
+ case TCG_CALL_ARG_EXTEND_U:
+ case TCG_CALL_ARG_EXTEND_S:
+ load_arg_normal_1(s, loc, ts, &allocated_regs);
+ break;
+ default:
+ g_assert_not_reached();
}
}
-
- /* mark dead temporaries and free the associated registers */
+
+ /* Mark dead temporaries and free the associated registers. */
for (i = nb_oargs; i < nb_iargs + nb_oargs; i++) {
if (IS_DEAD_ARG(i)) {
temp_dead(s, arg_temp(op->args[i]));
}
}
-
- /* clobber call registers */
+
+ /* Clobber call registers. */
for (i = 0; i < TCG_TARGET_NB_REGS; i++) {
if (tcg_regset_test_reg(tcg_target_call_clobber_regs, i)) {
tcg_reg_free(s, i, allocated_regs);
}
}
- /* Save globals if they might be written by the helper, sync them if
- they might be read. */
- if (flags & TCG_CALL_NO_READ_GLOBALS) {
+ /*
+ * Save globals if they might be written by the helper,
+ * sync them if they might be read.
+ */
+ if (info->flags & TCG_CALL_NO_READ_GLOBALS) {
/* Nothing to do */
- } else if (flags & TCG_CALL_NO_WRITE_GLOBALS) {
+ } else if (info->flags & TCG_CALL_NO_WRITE_GLOBALS) {
sync_globals(s, allocated_regs);
} else {
save_globals(s, allocated_regs);
@@ -4268,31 +4391,41 @@ static void tcg_reg_alloc_call(TCGContext *s, TCGOp *op)
gpointer hash = (gpointer)(uintptr_t)info->typemask;
ffi_cif *cif = g_hash_table_lookup(ffi_table, hash);
assert(cif != NULL);
- tcg_out_call(s, func_addr, cif);
+ tcg_out_call(s, tcg_call_func(op), cif);
}
#else
- tcg_out_call(s, func_addr);
+ tcg_out_call(s, tcg_call_func(op));
#endif
- /* assign output registers and emit moves if needed */
- for(i = 0; i < nb_oargs; i++) {
- arg = op->args[i];
- ts = arg_temp(arg);
+ /* Assign output registers and emit moves if needed. */
+ switch (info->out_kind) {
+ case TCG_CALL_RET_NORMAL:
+ for (i = 0; i < nb_oargs; i++) {
+ TCGTemp *ts = arg_temp(op->args[i]);
+ TCGReg reg = tcg_target_call_oarg_regs[i];
- /* ENV should not be modified. */
- tcg_debug_assert(!temp_readonly(ts));
+ /* ENV should not be modified. */
+ tcg_debug_assert(!temp_readonly(ts));
- reg = tcg_target_call_oarg_regs[i];
- tcg_debug_assert(s->reg_to_temp[reg] == NULL);
- if (ts->val_type == TEMP_VAL_REG) {
- s->reg_to_temp[ts->reg] = NULL;
+ tcg_debug_assert(s->reg_to_temp[reg] == NULL);
+ if (ts->val_type == TEMP_VAL_REG) {
+ s->reg_to_temp[ts->reg] = NULL;
+ }
+ ts->val_type = TEMP_VAL_REG;
+ ts->reg = reg;
+ ts->mem_coherent = 0;
+ s->reg_to_temp[reg] = ts;
}
- ts->val_type = TEMP_VAL_REG;
- ts->reg = reg;
- ts->mem_coherent = 0;
- s->reg_to_temp[reg] = ts;
+ break;
+ default:
+ g_assert_not_reached();
+ }
+
+ /* Flush or discard output registers as needed. */
+ for (i = 0; i < nb_oargs; i++) {
+ TCGTemp *ts = arg_temp(op->args[i]);
if (NEED_SYNC_ARG(i)) {
- temp_sync(s, ts, allocated_regs, 0, IS_DEAD_ARG(i));
+ temp_sync(s, ts, s->reserved_regs, 0, IS_DEAD_ARG(i));
} else if (IS_DEAD_ARG(i)) {
temp_dead(s, ts);
}
--
2.34.1
^ permalink raw reply related [flat|nested] 43+ messages in thread* [PATCH v2 21/36] tcg: Move ffi_cif pointer into TCGHelperInfo
2022-10-21 7:15 [PATCH v2 00/36] tcg: Support for Int128 with helpers Richard Henderson
` (19 preceding siblings ...)
2022-10-21 7:15 ` [PATCH v2 20/36] tcg: Reorg function calls Richard Henderson
@ 2022-10-21 7:15 ` Richard Henderson
2022-10-21 7:15 ` [PATCH v2 22/36] tcg: Add TCGHelperInfo argument to tcg_out_call Richard Henderson
` (14 subsequent siblings)
35 siblings, 0 replies; 43+ messages in thread
From: Richard Henderson @ 2022-10-21 7:15 UTC (permalink / raw)
To: qemu-devel
Instead of requiring a separate hash table lookup,
put a pointer to the CIF into TCGHelperInfo.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
tcg/tcg-internal.h | 7 +++
tcg/tcg.c | 129 +++++++++++++++++++++++++--------------------
2 files changed, 78 insertions(+), 58 deletions(-)
diff --git a/tcg/tcg-internal.h b/tcg/tcg-internal.h
index 097fef2325..696dc66ada 100644
--- a/tcg/tcg-internal.h
+++ b/tcg/tcg-internal.h
@@ -25,6 +25,10 @@
#ifndef TCG_INTERNAL_H
#define TCG_INTERNAL_H
+#ifdef CONFIG_TCG_INTERPRETER
+#include <ffi.h>
+#endif
+
#define TCG_HIGHWATER 1024
/*
@@ -54,6 +58,9 @@ typedef struct TCGCallArgumentLoc {
typedef struct TCGHelperInfo {
void *func;
const char *name;
+#ifdef CONFIG_TCG_INTERPRETER
+ ffi_cif *cif;
+#endif
unsigned typemask : 32;
unsigned flags : 8;
unsigned nr_in : 8;
diff --git a/tcg/tcg.c b/tcg/tcg.c
index 713e692621..a3a6968f76 100644
--- a/tcg/tcg.c
+++ b/tcg/tcg.c
@@ -62,10 +62,6 @@
#include "tcg/tcg-ldst.h"
#include "tcg-internal.h"
-#ifdef CONFIG_TCG_INTERPRETER
-#include <ffi.h>
-#endif
-
/* Forward declarations for functions declared in tcg-target.c.inc and
used here. */
static void tcg_target_init(TCGContext *s);
@@ -552,17 +548,76 @@ static TCGHelperInfo all_helpers[] = {
static GHashTable *helper_table;
#ifdef CONFIG_TCG_INTERPRETER
-static GHashTable *ffi_table;
+static ffi_type *typecode_to_ffi(int argmask)
+{
+ switch (argmask) {
+ case dh_typecode_void:
+ return &ffi_type_void;
+ case dh_typecode_i32:
+ return &ffi_type_uint32;
+ case dh_typecode_s32:
+ return &ffi_type_sint32;
+ case dh_typecode_i64:
+ return &ffi_type_uint64;
+ case dh_typecode_s64:
+ return &ffi_type_sint64;
+ case dh_typecode_ptr:
+ return &ffi_type_pointer;
+ }
+ g_assert_not_reached();
+}
-static ffi_type * const typecode_to_ffi[8] = {
- [dh_typecode_void] = &ffi_type_void,
- [dh_typecode_i32] = &ffi_type_uint32,
- [dh_typecode_s32] = &ffi_type_sint32,
- [dh_typecode_i64] = &ffi_type_uint64,
- [dh_typecode_s64] = &ffi_type_sint64,
- [dh_typecode_ptr] = &ffi_type_pointer,
-};
-#endif
+static void init_ffi_layouts(void)
+{
+ /* g_direct_hash/equal for direct comparisons on uint32_t. */
+ GHashTable *ffi_table = g_hash_table_new(NULL, NULL);
+
+ for (int i = 0; i < ARRAY_SIZE(all_helpers); ++i) {
+ TCGHelperInfo *info = &all_helpers[i];
+ unsigned typemask = info->typemask;
+ gpointer hash = (gpointer)(uintptr_t)typemask;
+ struct {
+ ffi_cif cif;
+ ffi_type *args[];
+ } *ca;
+ ffi_status status;
+ int nargs;
+ ffi_cif *cif;
+
+ cif = g_hash_table_lookup(ffi_table, hash);
+ if (cif) {
+ info->cif = cif;
+ continue;
+ }
+
+ /* Ignoring the return type, find the last non-zero field. */
+ nargs = 32 - clz32(typemask >> 3);
+ nargs = DIV_ROUND_UP(nargs, 3);
+
+ ca = g_malloc0(sizeof(*ca) + nargs * sizeof(ffi_type *));
+ ca->cif.rtype = typecode_to_ffi(typemask & 7);
+ ca->cif.nargs = nargs;
+
+ if (nargs != 0) {
+ ca->cif.arg_types = ca->args;
+ for (i = 0; i < nargs; ++i) {
+ int typecode = extract32(typemask, (i + 1) * 3, 3);
+ ca->args[i] = typecode_to_ffi(typecode);
+ }
+ }
+
+ status = ffi_prep_cif(&ca->cif, FFI_DEFAULT_ABI, nargs,
+ ca->cif.rtype, ca->cif.arg_types);
+ assert(status == FFI_OK);
+
+ cif = &ca->cif;
+ info->cif = cif;
+ g_hash_table_insert(ffi_table, hash, (gpointer)cif);
+ }
+
+ g_hash_table_destroy(ffi_table);
+}
+#endif /* CONFIG_TCG_INTERPRETER */
typedef struct TCGCumulativeArgs {
int arg_idx; /* tcg_gen_callN args[] */
@@ -765,44 +820,7 @@ static void tcg_context_init(unsigned max_cpus)
}
#ifdef CONFIG_TCG_INTERPRETER
- /* g_direct_hash/equal for direct comparisons on uint32_t. */
- ffi_table = g_hash_table_new(NULL, NULL);
- for (i = 0; i < ARRAY_SIZE(all_helpers); ++i) {
- struct {
- ffi_cif cif;
- ffi_type *args[];
- } *ca;
- uint32_t typemask = all_helpers[i].typemask;
- gpointer hash = (gpointer)(uintptr_t)typemask;
- ffi_status status;
- int nargs;
-
- if (g_hash_table_lookup(ffi_table, hash)) {
- continue;
- }
-
- /* Ignoring the return type, find the last non-zero field. */
- nargs = 32 - clz32(typemask >> 3);
- nargs = DIV_ROUND_UP(nargs, 3);
-
- ca = g_malloc0(sizeof(*ca) + nargs * sizeof(ffi_type *));
- ca->cif.rtype = typecode_to_ffi[typemask & 7];
- ca->cif.nargs = nargs;
-
- if (nargs != 0) {
- ca->cif.arg_types = ca->args;
- for (i = 0; i < nargs; ++i) {
- int typecode = extract32(typemask, (i + 1) * 3, 3);
- ca->args[i] = typecode_to_ffi[typecode];
- }
- }
-
- status = ffi_prep_cif(&ca->cif, FFI_DEFAULT_ABI, nargs,
- ca->cif.rtype, ca->cif.arg_types);
- assert(status == FFI_OK);
-
- g_hash_table_insert(ffi_table, hash, (gpointer)&ca->cif);
- }
+ init_ffi_layouts();
#endif
tcg_target_init(s);
@@ -4387,12 +4405,7 @@ static void tcg_reg_alloc_call(TCGContext *s, TCGOp *op)
}
#ifdef CONFIG_TCG_INTERPRETER
- {
- gpointer hash = (gpointer)(uintptr_t)info->typemask;
- ffi_cif *cif = g_hash_table_lookup(ffi_table, hash);
- assert(cif != NULL);
- tcg_out_call(s, tcg_call_func(op), cif);
- }
+ tcg_out_call(s, tcg_call_func(op), info->cif);
#else
tcg_out_call(s, tcg_call_func(op));
#endif
--
2.34.1
^ permalink raw reply related [flat|nested] 43+ messages in thread* [PATCH v2 22/36] tcg: Add TCGHelperInfo argument to tcg_out_call
2022-10-21 7:15 [PATCH v2 00/36] tcg: Support for Int128 with helpers Richard Henderson
` (20 preceding siblings ...)
2022-10-21 7:15 ` [PATCH v2 21/36] tcg: Move ffi_cif pointer into TCGHelperInfo Richard Henderson
@ 2022-10-21 7:15 ` Richard Henderson
2022-10-21 7:15 ` [PATCH v2 23/36] tcg: Define TCG_TYPE_I128 and related helper macros Richard Henderson
` (13 subsequent siblings)
35 siblings, 0 replies; 43+ messages in thread
From: Richard Henderson @ 2022-10-21 7:15 UTC (permalink / raw)
To: qemu-devel; +Cc: Philippe Mathieu-Daudé
This eliminates an ifdef for TCI, and will be required for
expanding the call for TCGv_i128.
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
tcg/tcg.c | 12 ++----------
tcg/aarch64/tcg-target.c.inc | 19 ++++++++++---------
tcg/arm/tcg-target.c.inc | 10 ++++++++--
tcg/i386/tcg-target.c.inc | 5 +++--
tcg/loongarch64/tcg-target.c.inc | 7 ++++---
tcg/mips/tcg-target.c.inc | 3 ++-
tcg/ppc/tcg-target.c.inc | 7 ++++---
tcg/riscv/tcg-target.c.inc | 7 ++++---
tcg/s390x/tcg-target.c.inc | 12 +++++++++---
tcg/sparc64/tcg-target.c.inc | 3 ++-
tcg/tci/tcg-target.c.inc | 3 ++-
11 files changed, 50 insertions(+), 38 deletions(-)
diff --git a/tcg/tcg.c b/tcg/tcg.c
index a3a6968f76..082482341b 100644
--- a/tcg/tcg.c
+++ b/tcg/tcg.c
@@ -145,12 +145,8 @@ static void tcg_out_st(TCGContext *s, TCGType type, TCGReg arg, TCGReg arg1,
intptr_t arg2);
static bool tcg_out_sti(TCGContext *s, TCGType type, TCGArg val,
TCGReg base, intptr_t ofs);
-#ifdef CONFIG_TCG_INTERPRETER
static void tcg_out_call(TCGContext *s, const tcg_insn_unit *target,
- ffi_cif *cif);
-#else
-static void tcg_out_call(TCGContext *s, const tcg_insn_unit *target);
-#endif
+ const TCGHelperInfo *info);
static bool tcg_target_const_match(int64_t val, TCGType type, int ct);
#ifdef TCG_TARGET_NEED_LDST_LABELS
static int tcg_out_ldst_finalize(TCGContext *s);
@@ -4404,11 +4400,7 @@ static void tcg_reg_alloc_call(TCGContext *s, TCGOp *op)
save_globals(s, allocated_regs);
}
-#ifdef CONFIG_TCG_INTERPRETER
- tcg_out_call(s, tcg_call_func(op), info->cif);
-#else
- tcg_out_call(s, tcg_call_func(op));
-#endif
+ tcg_out_call(s, tcg_call_func(op), info);
/* Assign output registers and emit moves if needed. */
switch (info->out_kind) {
diff --git a/tcg/aarch64/tcg-target.c.inc b/tcg/aarch64/tcg-target.c.inc
index d997f7922a..e8184fe001 100644
--- a/tcg/aarch64/tcg-target.c.inc
+++ b/tcg/aarch64/tcg-target.c.inc
@@ -1336,22 +1336,23 @@ static void tcg_out_goto_long(TCGContext *s, const tcg_insn_unit *target)
}
}
-static inline void tcg_out_callr(TCGContext *s, TCGReg reg)
-{
- tcg_out_insn(s, 3207, BLR, reg);
-}
-
-static void tcg_out_call(TCGContext *s, const tcg_insn_unit *target)
+static void tcg_out_call_int(TCGContext *s, const tcg_insn_unit *target)
{
ptrdiff_t offset = tcg_pcrel_diff(s, target) >> 2;
if (offset == sextract64(offset, 0, 26)) {
tcg_out_insn(s, 3206, BL, offset);
} else {
tcg_out_movi(s, TCG_TYPE_I64, TCG_REG_TMP, (intptr_t)target);
- tcg_out_callr(s, TCG_REG_TMP);
+ tcg_out_insn(s, 3207, BLR, reg);
}
}
+static void tcg_out_call(TCGContext *s, const tcg_insn_unit *target,
+ const TCGHelperInfo *info)
+{
+ tcg_out_call_int(s, target);
+}
+
void tb_target_set_jmp_target(uintptr_t tc_ptr, uintptr_t jmp_rx,
uintptr_t jmp_rw, uintptr_t addr)
{
@@ -1599,7 +1600,7 @@ static bool tcg_out_qemu_ld_slow_path(TCGContext *s, TCGLabelQemuLdst *lb)
tcg_out_mov(s, TARGET_LONG_BITS == 64, TCG_REG_X1, lb->addrlo_reg);
tcg_out_movi(s, TCG_TYPE_I32, TCG_REG_X2, oi);
tcg_out_adr(s, TCG_REG_X3, lb->raddr);
- tcg_out_call(s, qemu_ld_helpers[opc & MO_SIZE]);
+ tcg_out_call_int(s, qemu_ld_helpers[opc & MO_SIZE]);
if (opc & MO_SIGN) {
tcg_out_sxt(s, lb->type, size, lb->datalo_reg, TCG_REG_X0);
} else {
@@ -1625,7 +1626,7 @@ static bool tcg_out_qemu_st_slow_path(TCGContext *s, TCGLabelQemuLdst *lb)
tcg_out_mov(s, size == MO_64, TCG_REG_X2, lb->datalo_reg);
tcg_out_movi(s, TCG_TYPE_I32, TCG_REG_X3, oi);
tcg_out_adr(s, TCG_REG_X4, lb->raddr);
- tcg_out_call(s, qemu_st_helpers[opc & MO_SIZE]);
+ tcg_out_call_int(s, qemu_st_helpers[opc & MO_SIZE]);
tcg_out_goto(s, lb->raddr);
return true;
}
diff --git a/tcg/arm/tcg-target.c.inc b/tcg/arm/tcg-target.c.inc
index aa3a888fed..e82749a602 100644
--- a/tcg/arm/tcg-target.c.inc
+++ b/tcg/arm/tcg-target.c.inc
@@ -1131,7 +1131,7 @@ static void tcg_out_goto(TCGContext *s, ARMCond cond, const tcg_insn_unit *addr)
* The call case is mostly used for helpers - so it's not unreasonable
* for them to be beyond branch range.
*/
-static void tcg_out_call(TCGContext *s, const tcg_insn_unit *addr)
+static void tcg_out_call_int(TCGContext *s, const tcg_insn_unit *addr)
{
intptr_t addri = (intptr_t)addr;
ptrdiff_t disp = tcg_pcrel_diff(s, addr);
@@ -1150,6 +1150,12 @@ static void tcg_out_call(TCGContext *s, const tcg_insn_unit *addr)
tcg_out_blx_reg(s, COND_AL, TCG_REG_TMP);
}
+static void tcg_out_call(TCGContext *s, const tcg_insn_unit *addr,
+ const TCGHelperInfo *info)
+{
+ tcg_out_call_int(s, addr);
+}
+
static void tcg_out_goto_label(TCGContext *s, ARMCond cond, TCGLabel *l)
{
if (l->has_value) {
@@ -1515,7 +1521,7 @@ static bool tcg_out_qemu_ld_slow_path(TCGContext *s, TCGLabelQemuLdst *lb)
argreg = tcg_out_arg_reg32(s, argreg, TCG_REG_R14);
/* Use the canonical unsigned helpers and minimize icache usage. */
- tcg_out_call(s, qemu_ld_helpers[opc & MO_SIZE]);
+ tcg_out_call_int(s, qemu_ld_helpers[opc & MO_SIZE]);
datalo = lb->datalo_reg;
datahi = lb->datahi_reg;
diff --git a/tcg/i386/tcg-target.c.inc b/tcg/i386/tcg-target.c.inc
index cb04e4b3ad..58bd5873f5 100644
--- a/tcg/i386/tcg-target.c.inc
+++ b/tcg/i386/tcg-target.c.inc
@@ -1661,7 +1661,8 @@ static void tcg_out_branch(TCGContext *s, int call, const tcg_insn_unit *dest)
}
}
-static inline void tcg_out_call(TCGContext *s, const tcg_insn_unit *dest)
+static void tcg_out_call(TCGContext *s, const tcg_insn_unit *dest,
+ const TCGHelperInfo *info)
{
tcg_out_branch(s, 1, dest);
}
@@ -1885,7 +1886,7 @@ static bool tcg_out_qemu_ld_slow_path(TCGContext *s, TCGLabelQemuLdst *l)
(uintptr_t)l->raddr);
}
- tcg_out_call(s, qemu_ld_helpers[opc & (MO_BSWAP | MO_SIZE)]);
+ tcg_out_branch(s, 1, qemu_ld_helpers[opc & (MO_BSWAP | MO_SIZE)]);
data_reg = l->datalo_reg;
switch (opc & MO_SSIZE) {
diff --git a/tcg/loongarch64/tcg-target.c.inc b/tcg/loongarch64/tcg-target.c.inc
index a3debf6da7..98da264f16 100644
--- a/tcg/loongarch64/tcg-target.c.inc
+++ b/tcg/loongarch64/tcg-target.c.inc
@@ -567,7 +567,8 @@ static void tcg_out_call_int(TCGContext *s, const tcg_insn_unit *arg, bool tail)
}
}
-static void tcg_out_call(TCGContext *s, const tcg_insn_unit *arg)
+static void tcg_out_call(TCGContext *s, const tcg_insn_unit *arg,
+ const TCGHelperInfo *info)
{
tcg_out_call_int(s, arg, false);
}
@@ -760,7 +761,7 @@ static bool tcg_out_qemu_ld_slow_path(TCGContext *s, TCGLabelQemuLdst *l)
tcg_out_movi(s, TCG_TYPE_PTR, TCG_REG_A2, oi);
tcg_out_movi(s, TCG_TYPE_PTR, TCG_REG_A3, (tcg_target_long)l->raddr);
- tcg_out_call(s, qemu_ld_helpers[size]);
+ tcg_out_call_int(s, qemu_ld_helpers[size], false);
switch (opc & MO_SSIZE) {
case MO_SB:
@@ -821,7 +822,7 @@ static bool tcg_out_qemu_st_slow_path(TCGContext *s, TCGLabelQemuLdst *l)
tcg_out_movi(s, TCG_TYPE_PTR, TCG_REG_A3, oi);
tcg_out_movi(s, TCG_TYPE_PTR, TCG_REG_A4, (tcg_target_long)l->raddr);
- tcg_out_call(s, qemu_st_helpers[size]);
+ tcg_out_call_int(s, qemu_st_helpers[size], false);
return tcg_out_goto(s, l->raddr);
}
diff --git a/tcg/mips/tcg-target.c.inc b/tcg/mips/tcg-target.c.inc
index bd76f0c97f..292e490b5c 100644
--- a/tcg/mips/tcg-target.c.inc
+++ b/tcg/mips/tcg-target.c.inc
@@ -1020,7 +1020,8 @@ static void tcg_out_call_int(TCGContext *s, const tcg_insn_unit *arg, bool tail)
}
}
-static void tcg_out_call(TCGContext *s, const tcg_insn_unit *arg)
+static void tcg_out_call(TCGContext *s, const tcg_insn_unit *arg,
+ const TCGHelperInfo *info)
{
tcg_out_call_int(s, arg, false);
tcg_out_nop(s);
diff --git a/tcg/ppc/tcg-target.c.inc b/tcg/ppc/tcg-target.c.inc
index 500b75a03a..f561a3492f 100644
--- a/tcg/ppc/tcg-target.c.inc
+++ b/tcg/ppc/tcg-target.c.inc
@@ -2002,7 +2002,8 @@ static void tcg_out_call_int(TCGContext *s, int lk,
#endif
}
-static void tcg_out_call(TCGContext *s, const tcg_insn_unit *target)
+static void tcg_out_call(TCGContext *s, const tcg_insn_unit *target,
+ const TCGHelperInfo *info)
{
tcg_out_call_int(s, LK, target);
}
@@ -2221,7 +2222,7 @@ static bool tcg_out_qemu_ld_slow_path(TCGContext *s, TCGLabelQemuLdst *lb)
tcg_out_movi(s, TCG_TYPE_I32, arg++, oi);
tcg_out32(s, MFSPR | RT(arg) | LR);
- tcg_out_call(s, qemu_ld_helpers[opc & (MO_BSWAP | MO_SIZE)]);
+ tcg_out_call_int(s, LK, qemu_ld_helpers[opc & (MO_BSWAP | MO_SIZE)]);
lo = lb->datalo_reg;
hi = lb->datahi_reg;
@@ -2290,7 +2291,7 @@ static bool tcg_out_qemu_st_slow_path(TCGContext *s, TCGLabelQemuLdst *lb)
tcg_out_movi(s, TCG_TYPE_I32, arg++, oi);
tcg_out32(s, MFSPR | RT(arg) | LR);
- tcg_out_call(s, qemu_st_helpers[opc & (MO_BSWAP | MO_SIZE)]);
+ tcg_out_call_int(s, LK, qemu_st_helpers[opc & (MO_BSWAP | MO_SIZE)]);
tcg_out_b(s, 0, lb->raddr);
return true;
diff --git a/tcg/riscv/tcg-target.c.inc b/tcg/riscv/tcg-target.c.inc
index 81a83e45b1..aa017d665a 100644
--- a/tcg/riscv/tcg-target.c.inc
+++ b/tcg/riscv/tcg-target.c.inc
@@ -819,7 +819,8 @@ static void tcg_out_call_int(TCGContext *s, const tcg_insn_unit *arg, bool tail)
}
}
-static void tcg_out_call(TCGContext *s, const tcg_insn_unit *arg)
+static void tcg_out_call(TCGContext *s, const tcg_insn_unit *arg,
+ const TCGHelperInfo *info)
{
tcg_out_call_int(s, arg, false);
}
@@ -1002,7 +1003,7 @@ static bool tcg_out_qemu_ld_slow_path(TCGContext *s, TCGLabelQemuLdst *l)
tcg_out_movi(s, TCG_TYPE_PTR, a2, oi);
tcg_out_movi(s, TCG_TYPE_PTR, a3, (tcg_target_long)l->raddr);
- tcg_out_call(s, qemu_ld_helpers[opc & MO_SSIZE]);
+ tcg_out_call_int(s, qemu_ld_helpers[opc & MO_SSIZE], false);
tcg_out_mov(s, (opc & MO_SIZE) == MO_64, l->datalo_reg, a0);
tcg_out_goto(s, l->raddr);
@@ -1047,7 +1048,7 @@ static bool tcg_out_qemu_st_slow_path(TCGContext *s, TCGLabelQemuLdst *l)
tcg_out_movi(s, TCG_TYPE_PTR, a3, oi);
tcg_out_movi(s, TCG_TYPE_PTR, a4, (tcg_target_long)l->raddr);
- tcg_out_call(s, qemu_st_helpers[opc & MO_SIZE]);
+ tcg_out_call_int(s, qemu_st_helpers[opc & MO_SIZE], false);
tcg_out_goto(s, l->raddr);
return true;
diff --git a/tcg/s390x/tcg-target.c.inc b/tcg/s390x/tcg-target.c.inc
index 47aaba7667..c3c0bcc3eb 100644
--- a/tcg/s390x/tcg-target.c.inc
+++ b/tcg/s390x/tcg-target.c.inc
@@ -1691,7 +1691,7 @@ static void tgen_brcond(TCGContext *s, TCGType type, TCGCond c,
tgen_branch(s, cc, l);
}
-static void tcg_out_call(TCGContext *s, const tcg_insn_unit *dest)
+static void tcg_out_call_int(TCGContext *s, const tcg_insn_unit *dest)
{
ptrdiff_t off = tcg_pcrel_diff(s, dest) >> 1;
if (off == (int32_t)off) {
@@ -1702,6 +1702,12 @@ static void tcg_out_call(TCGContext *s, const tcg_insn_unit *dest)
}
}
+static void tcg_out_call(TCGContext *s, const tcg_insn_unit *dest,
+ const TCGHelperInfo *info)
+{
+ tcg_out_call_int(s, dest);
+}
+
static void tcg_out_qemu_ld_direct(TCGContext *s, MemOp opc, TCGReg data,
TCGReg base, TCGReg index, int disp)
{
@@ -1897,7 +1903,7 @@ static bool tcg_out_qemu_ld_slow_path(TCGContext *s, TCGLabelQemuLdst *lb)
}
tcg_out_movi(s, TCG_TYPE_I32, TCG_REG_R4, oi);
tcg_out_movi(s, TCG_TYPE_PTR, TCG_REG_R5, (uintptr_t)lb->raddr);
- tcg_out_call(s, qemu_ld_helpers[opc & (MO_BSWAP | MO_SSIZE)]);
+ tcg_out_call_int(s, qemu_ld_helpers[opc & (MO_BSWAP | MO_SSIZE)]);
tcg_out_mov(s, TCG_TYPE_I64, data_reg, TCG_REG_R2);
tgen_gotoi(s, S390_CC_ALWAYS, lb->raddr);
@@ -1938,7 +1944,7 @@ static bool tcg_out_qemu_st_slow_path(TCGContext *s, TCGLabelQemuLdst *lb)
}
tcg_out_movi(s, TCG_TYPE_I32, TCG_REG_R5, oi);
tcg_out_movi(s, TCG_TYPE_PTR, TCG_REG_R6, (uintptr_t)lb->raddr);
- tcg_out_call(s, qemu_st_helpers[opc & (MO_BSWAP | MO_SIZE)]);
+ tcg_out_call_int(s, qemu_st_helpers[opc & (MO_BSWAP | MO_SIZE)]);
tgen_gotoi(s, S390_CC_ALWAYS, lb->raddr);
return true;
diff --git a/tcg/sparc64/tcg-target.c.inc b/tcg/sparc64/tcg-target.c.inc
index cb9453efdd..eb913f33c8 100644
--- a/tcg/sparc64/tcg-target.c.inc
+++ b/tcg/sparc64/tcg-target.c.inc
@@ -859,7 +859,8 @@ static void tcg_out_call_nodelay(TCGContext *s, const tcg_insn_unit *dest,
}
}
-static void tcg_out_call(TCGContext *s, const tcg_insn_unit *dest)
+static void tcg_out_call(TCGContext *s, const tcg_insn_unit *dest,
+ const TCGHelperInfo *info)
{
tcg_out_call_nodelay(s, dest, false);
tcg_out_nop(s);
diff --git a/tcg/tci/tcg-target.c.inc b/tcg/tci/tcg-target.c.inc
index f3d7441e06..b07b30b9fc 100644
--- a/tcg/tci/tcg-target.c.inc
+++ b/tcg/tci/tcg-target.c.inc
@@ -562,8 +562,9 @@ static void tcg_out_movi(TCGContext *s, TCGType type,
}
static void tcg_out_call(TCGContext *s, const tcg_insn_unit *func,
- ffi_cif *cif)
+ const TCGHelperInfo *info)
{
+ ffi_cif *cif = info->cif;
tcg_insn_unit insn = 0;
uint8_t which;
--
2.34.1
^ permalink raw reply related [flat|nested] 43+ messages in thread* [PATCH v2 23/36] tcg: Define TCG_TYPE_I128 and related helper macros
2022-10-21 7:15 [PATCH v2 00/36] tcg: Support for Int128 with helpers Richard Henderson
` (21 preceding siblings ...)
2022-10-21 7:15 ` [PATCH v2 22/36] tcg: Add TCGHelperInfo argument to tcg_out_call Richard Henderson
@ 2022-10-21 7:15 ` Richard Henderson
2022-10-21 7:15 ` [PATCH v2 24/36] tcg: Add TCG_CALL_{RET,ARG}_NORMAL_4 Richard Henderson
` (12 subsequent siblings)
35 siblings, 0 replies; 43+ messages in thread
From: Richard Henderson @ 2022-10-21 7:15 UTC (permalink / raw)
To: qemu-devel
Begin staging in support for TCGv_i128 with Int128.
Define the type enumerator, the typedef, and the
helper-head.h macros.
This cannot yet be used, because you can't allocate
temporaries of this new type.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
include/exec/helper-head.h | 7 +++++++
include/tcg/tcg.h | 28 ++++++++++++++--------------
2 files changed, 21 insertions(+), 14 deletions(-)
diff --git a/include/exec/helper-head.h b/include/exec/helper-head.h
index e242fed46e..9753df8e80 100644
--- a/include/exec/helper-head.h
+++ b/include/exec/helper-head.h
@@ -26,6 +26,7 @@
#define dh_alias_int i32
#define dh_alias_i64 i64
#define dh_alias_s64 i64
+#define dh_alias_i128 i128
#define dh_alias_f16 i32
#define dh_alias_f32 i32
#define dh_alias_f64 i64
@@ -40,6 +41,7 @@
#define dh_ctype_int int
#define dh_ctype_i64 uint64_t
#define dh_ctype_s64 int64_t
+#define dh_ctype_i128 Int128
#define dh_ctype_f16 uint32_t
#define dh_ctype_f32 float32
#define dh_ctype_f64 float64
@@ -71,6 +73,7 @@
#define dh_retvar_decl0_noreturn void
#define dh_retvar_decl0_i32 TCGv_i32 retval
#define dh_retvar_decl0_i64 TCGv_i64 retval
+#define dh_retval_decl0_i128 TCGv_i128 retval
#define dh_retvar_decl0_ptr TCGv_ptr retval
#define dh_retvar_decl0(t) glue(dh_retvar_decl0_, dh_alias(t))
@@ -78,6 +81,7 @@
#define dh_retvar_decl_noreturn
#define dh_retvar_decl_i32 TCGv_i32 retval,
#define dh_retvar_decl_i64 TCGv_i64 retval,
+#define dh_retvar_decl_i128 TCGv_i128 retval,
#define dh_retvar_decl_ptr TCGv_ptr retval,
#define dh_retvar_decl(t) glue(dh_retvar_decl_, dh_alias(t))
@@ -85,6 +89,7 @@
#define dh_retvar_noreturn NULL
#define dh_retvar_i32 tcgv_i32_temp(retval)
#define dh_retvar_i64 tcgv_i64_temp(retval)
+#define dh_retvar_i128 tcgv_i128_temp(retval)
#define dh_retvar_ptr tcgv_ptr_temp(retval)
#define dh_retvar(t) glue(dh_retvar_, dh_alias(t))
@@ -95,6 +100,7 @@
#define dh_typecode_i64 4
#define dh_typecode_s64 5
#define dh_typecode_ptr 6
+#define dh_typecode_i128 7
#define dh_typecode_int dh_typecode_s32
#define dh_typecode_f16 dh_typecode_i32
#define dh_typecode_f32 dh_typecode_i32
@@ -104,6 +110,7 @@
#define dh_callflag_i32 0
#define dh_callflag_i64 0
+#define dh_callflag_i128 0
#define dh_callflag_ptr 0
#define dh_callflag_void 0
#define dh_callflag_noreturn TCG_CALL_NO_RETURN
diff --git a/include/tcg/tcg.h b/include/tcg/tcg.h
index 8d0626c797..88fb4f1d17 100644
--- a/include/tcg/tcg.h
+++ b/include/tcg/tcg.h
@@ -38,18 +38,15 @@
/* XXX: make safe guess about sizes */
#define MAX_OP_PER_INSTR 266
-#if HOST_LONG_BITS == 32
#define MAX_OPC_PARAM_PER_ARG 2
-#else
-#define MAX_OPC_PARAM_PER_ARG 1
-#endif
#define MAX_OPC_PARAM_IARGS 7
#define MAX_OPC_PARAM_OARGS 1
#define MAX_OPC_PARAM_ARGS (MAX_OPC_PARAM_IARGS + MAX_OPC_PARAM_OARGS)
-/* A Call op needs up to 4 + 2N parameters on 32-bit archs,
- * and up to 4 + N parameters on 64-bit archs
- * (N = number of input arguments + output arguments). */
+/*
+ * A Call op needs up to 4 + 2N parameters.
+ * (N = number of input arguments + output arguments).
+ */
#define MAX_OPC_PARAM (4 + (MAX_OPC_PARAM_PER_ARG * MAX_OPC_PARAM_ARGS))
#define CPU_TEMP_BUF_NLONGS 128
@@ -289,6 +286,7 @@ typedef struct TCGPool {
typedef enum TCGType {
TCG_TYPE_I32,
TCG_TYPE_I64,
+ TCG_TYPE_I128,
TCG_TYPE_V64,
TCG_TYPE_V128,
@@ -370,13 +368,14 @@ typedef tcg_target_ulong TCGArg;
in tcg/README. Target CPU front-end code uses these types to deal
with TCG variables as it emits TCG code via the tcg_gen_* functions.
They come in several flavours:
- * TCGv_i32 : 32 bit integer type
- * TCGv_i64 : 64 bit integer type
- * TCGv_ptr : a host pointer type
- * TCGv_vec : a host vector type; the exact size is not exposed
- to the CPU front-end code.
- * TCGv : an integer type the same size as target_ulong
- (an alias for either TCGv_i32 or TCGv_i64)
+ * TCGv_i32 : 32 bit integer type
+ * TCGv_i64 : 64 bit integer type
+ * TCGv_i128 : 128 bit integer type
+ * TCGv_ptr : a host pointer type
+ * TCGv_vec : a host vector type; the exact size is not exposed
+ to the CPU front-end code.
+ * TCGv : an integer type the same size as target_ulong
+ (an alias for either TCGv_i32 or TCGv_i64)
The compiler's type checking will complain if you mix them
up and pass the wrong sized TCGv to a function.
@@ -396,6 +395,7 @@ typedef tcg_target_ulong TCGArg;
typedef struct TCGv_i32_d *TCGv_i32;
typedef struct TCGv_i64_d *TCGv_i64;
+typedef struct TCGv_i128_d *TCGv_i128;
typedef struct TCGv_ptr_d *TCGv_ptr;
typedef struct TCGv_vec_d *TCGv_vec;
typedef TCGv_ptr TCGv_env;
--
2.34.1
^ permalink raw reply related [flat|nested] 43+ messages in thread* [PATCH v2 24/36] tcg: Add TCG_CALL_{RET,ARG}_NORMAL_4
2022-10-21 7:15 [PATCH v2 00/36] tcg: Support for Int128 with helpers Richard Henderson
` (22 preceding siblings ...)
2022-10-21 7:15 ` [PATCH v2 23/36] tcg: Define TCG_TYPE_I128 and related helper macros Richard Henderson
@ 2022-10-21 7:15 ` Richard Henderson
2022-10-21 7:15 ` [PATCH v2 25/36] tcg: Allocate objects contiguously in temp_allocate_frame Richard Henderson
` (11 subsequent siblings)
35 siblings, 0 replies; 43+ messages in thread
From: Richard Henderson @ 2022-10-21 7:15 UTC (permalink / raw)
To: qemu-devel
These will be used by some 32-bit hosts to pass and return i128.
Not yet used, because allocation is not yet enabled.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
tcg/tcg-internal.h | 2 +
tcg/tcg.c | 127 ++++++++++++++++++++++++++++++++++++++++++++-
2 files changed, 128 insertions(+), 1 deletion(-)
diff --git a/tcg/tcg-internal.h b/tcg/tcg-internal.h
index 696dc66ada..7d4f7049be 100644
--- a/tcg/tcg-internal.h
+++ b/tcg/tcg-internal.h
@@ -36,6 +36,7 @@
*/
typedef enum {
TCG_CALL_RET_NORMAL, /* by registers */
+ TCG_CALL_RET_NORMAL_4, /* for i128, by 4 registers */
} TCGCallReturnKind;
typedef enum {
@@ -44,6 +45,7 @@ typedef enum {
TCG_CALL_ARG_EXTEND, /* for i32, as a sign/zero-extended i64 */
TCG_CALL_ARG_EXTEND_U, /* ... as a zero-extended i64 */
TCG_CALL_ARG_EXTEND_S, /* ... as a sign-extended i64 */
+ TCG_CALL_ARG_NORMAL_4, /* for i128, like normal with 4 slots */
} TCGCallArgumentKind;
typedef struct TCGCallArgumentLoc {
diff --git a/tcg/tcg.c b/tcg/tcg.c
index 082482341b..fdafb30579 100644
--- a/tcg/tcg.c
+++ b/tcg/tcg.c
@@ -658,6 +658,30 @@ static void layout_arg_normal_2(TCGCumulativeArgs *cum, TCGHelperInfo *info)
loc[1].tmp_subindex = 1;
}
+static void layout_arg_normal_4(TCGCumulativeArgs *cum, TCGHelperInfo *info)
+{
+ TCGCallArgumentLoc *loc = &info->in[cum->info_in_idx];
+ int reg_n = MIN(4, cum->max_reg_slot - cum->reg_slot);
+
+ *loc = (TCGCallArgumentLoc){
+ .kind = TCG_CALL_ARG_NORMAL_4,
+ .arg_idx = cum->arg_idx,
+ };
+
+ if (reg_n > 0) {
+ loc->reg_slot = cum->reg_slot;
+ loc->reg_n = reg_n;
+ cum->reg_slot += reg_n;
+ }
+ if (reg_n < 4) {
+ loc->stk_slot = cum->stk_slot;
+ cum->stk_slot += 4 - reg_n;
+ }
+
+ cum->info_in_idx++;
+ cum->op_arg_idx++;
+}
+
static void init_call_layout(TCGHelperInfo *info)
{
unsigned typemask = info->typemask;
@@ -686,10 +710,29 @@ static void init_call_layout(TCGHelperInfo *info)
info->nr_out = 64 / TCG_TARGET_REG_BITS;
info->out_kind = TCG_CALL_RET_NORMAL;
break;
+ case dh_typecode_i128:
+ /*
+ * No matter the call return method, we must have all of
+ * the temp subindexes in the call for liveness.
+ */
+ info->nr_out = TCG_TARGET_REG_BITS == 32 ? 1 : 2;
+ info->out_kind = TCG_CALL_RET_NORMAL; /* TODO */
+ switch (/* TODO */ TCG_CALL_RET_NORMAL) {
+ case TCG_CALL_RET_NORMAL:
+ if (TCG_TARGET_REG_BITS == 32) {
+ assert(ARRAY_SIZE(tcg_target_call_oarg_regs) >= 4);
+ info->out_kind = TCG_CALL_RET_NORMAL_4;
+ } else {
+ assert(ARRAY_SIZE(tcg_target_call_oarg_regs) >= 2);
+ }
+ break;
+ default:
+ g_assert_not_reached();
+ }
+ break;
default:
g_assert_not_reached();
}
- assert(info->nr_out <= ARRAY_SIZE(tcg_target_call_oarg_regs));
/*
* The final two op->arg[] indexes are used for func & info.
@@ -745,6 +788,13 @@ static void init_call_layout(TCGHelperInfo *info)
layout_arg_1(&cum, info, TCG_CALL_ARG_NORMAL);
}
break;
+ case TCG_TYPE_I128:
+ if (TCG_TARGET_REG_BITS == 32) {
+ layout_arg_normal_4(&cum, info);
+ } else {
+ layout_arg_normal_2(&cum, info);
+ }
+ break;
default:
g_assert_not_reached();
}
@@ -1687,6 +1737,7 @@ void tcg_gen_callN(void *func, TCGTemp *ret, int nargs, TCGTemp **args)
switch (loc->kind) {
case TCG_CALL_ARG_NORMAL:
+ case TCG_CALL_ARG_NORMAL_4:
op->args[pi++] = temp_arg(ts);
break;
@@ -4301,6 +4352,41 @@ static bool tcg_reg_alloc_dup2(TCGContext *s, const TCGOp *op)
return true;
}
+static void copy_to_stk_i128(TCGContext *s, int stk_slot, TCGTemp *ts,
+ int slot, TCGRegSet allocated_regs)
+{
+ int stk_off = TCG_TARGET_CALL_STACK_OFFSET
+ + stk_slot * sizeof(tcg_target_long);
+
+ if (TCG_TARGET_REG_BITS == 32) {
+ TCGReg scratch;
+
+ tcg_debug_assert(ts->type == TCG_TYPE_I128);
+ tcg_debug_assert(ts->val_type == TEMP_VAL_MEM);
+ tcg_debug_assert(ts->mem_allocated);
+
+ scratch = tcg_reg_alloc(s, tcg_target_available_regs[TCG_TYPE_I32],
+ allocated_regs, 0, false);
+
+ for (; slot < 4; slot++) {
+ tcg_out_ld(s, TCG_TYPE_I32, scratch,
+ ts->mem_base->reg, ts->mem_offset + slot * 4);
+ tcg_out_st(s, TCG_TYPE_I32, scratch,
+ TCG_REG_CALL_STACK, stk_off + slot * 4);
+ }
+ } else {
+ tcg_debug_assert(ts->base_type == TCG_TYPE_I128);
+ tcg_debug_assert(ts->temp_subindex == 0);
+
+ for (; slot < 2; slot++) {
+ temp_load(s, &ts[slot], tcg_target_available_regs[TCG_TYPE_I64],
+ allocated_regs, 0);
+ tcg_out_st(s, TCG_TYPE_I64, ts[slot].reg,
+ TCG_REG_CALL_STACK, stk_off + slot * 8);
+ }
+ }
+}
+
static void load_arg_normal_1(TCGContext *s, const TCGCallArgumentLoc *loc,
TCGTemp *ts, TCGRegSet *allocated_regs)
{
@@ -4346,6 +4432,24 @@ static void load_arg_normal_1(TCGContext *s, const TCGCallArgumentLoc *loc,
tcg_regset_set_reg(*allocated_regs, reg);
}
+static void load_arg_normal_4(TCGContext *s, const TCGCallArgumentLoc *loc,
+ TCGTemp *ts, TCGRegSet *allocated_regs)
+{
+ int reg_n = loc->reg_n;
+
+ if (reg_n != 4) {
+ copy_to_stk_i128(s, loc->stk_slot, ts, reg_n, *allocated_regs);
+ }
+
+ for (reg_n--; reg_n >= 0; reg_n--) {
+ TCGReg reg = tcg_target_call_iarg_regs[loc->reg_slot + reg_n];
+
+ tcg_out_ld(s, TCG_TYPE_I32, reg,
+ ts->mem_base->reg, ts->mem_offset + reg_n * 4);
+ tcg_regset_set_reg(*allocated_regs, reg);
+ }
+}
+
static void tcg_reg_alloc_call(TCGContext *s, TCGOp *op)
{
const int nb_oargs = TCGOP_CALLO(op);
@@ -4369,6 +4473,9 @@ static void tcg_reg_alloc_call(TCGContext *s, TCGOp *op)
case TCG_CALL_ARG_EXTEND_S:
load_arg_normal_1(s, loc, ts, &allocated_regs);
break;
+ case TCG_CALL_ARG_NORMAL_4:
+ load_arg_normal_4(s, loc, ts, &allocated_regs);
+ break;
default:
g_assert_not_reached();
}
@@ -4422,6 +4529,24 @@ static void tcg_reg_alloc_call(TCGContext *s, TCGOp *op)
s->reg_to_temp[reg] = ts;
}
break;
+
+ case TCG_CALL_RET_NORMAL_4:
+ tcg_debug_assert(TCG_TARGET_REG_BITS == 32);
+ {
+ TCGTemp *ts = arg_temp(op->args[0]);
+
+ tcg_debug_assert(ts->type == TCG_TYPE_I128);
+ if (!ts->mem_allocated) {
+ temp_allocate_frame(s, ts);
+ }
+ for (i = 0; i < 4; i++) {
+ tcg_out_st(s, TCG_TYPE_I32, tcg_target_call_oarg_regs[i],
+ ts->mem_base->reg, ts->mem_offset + i * 4);
+ }
+ ts->val_type = TEMP_VAL_MEM;
+ }
+ break;
+
default:
g_assert_not_reached();
}
--
2.34.1
^ permalink raw reply related [flat|nested] 43+ messages in thread* [PATCH v2 25/36] tcg: Allocate objects contiguously in temp_allocate_frame
2022-10-21 7:15 [PATCH v2 00/36] tcg: Support for Int128 with helpers Richard Henderson
` (23 preceding siblings ...)
2022-10-21 7:15 ` [PATCH v2 24/36] tcg: Add TCG_CALL_{RET,ARG}_NORMAL_4 Richard Henderson
@ 2022-10-21 7:15 ` Richard Henderson
2022-10-21 7:15 ` [PATCH v2 26/36] tcg: Introduce tcg_out_addi_ptr Richard Henderson
` (10 subsequent siblings)
35 siblings, 0 replies; 43+ messages in thread
From: Richard Henderson @ 2022-10-21 7:15 UTC (permalink / raw)
To: qemu-devel
When allocating a temp to the stack frame, consider the
base type and allocate all parts at once.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
tcg/tcg.c | 30 ++++++++++++++++++++++--------
1 file changed, 22 insertions(+), 8 deletions(-)
diff --git a/tcg/tcg.c b/tcg/tcg.c
index fdafb30579..a3538edffa 100644
--- a/tcg/tcg.c
+++ b/tcg/tcg.c
@@ -3318,11 +3318,12 @@ static void check_regs(TCGContext *s)
static void temp_allocate_frame(TCGContext *s, TCGTemp *ts)
{
- int size = tcg_type_size(ts->type);
- int align;
intptr_t off;
+ int size, align;
- switch (ts->type) {
+ /* When allocating an object, look at the full type. */
+ size = tcg_type_size(ts->base_type);
+ switch (ts->base_type) {
case TCG_TYPE_I32:
align = 4;
break;
@@ -3353,13 +3354,26 @@ static void temp_allocate_frame(TCGContext *s, TCGTemp *ts)
tcg_raise_tb_overflow(s);
}
s->current_frame_offset = off + size;
-
- ts->mem_offset = off;
#if defined(__sparc__)
- ts->mem_offset += TCG_TARGET_STACK_BIAS;
+ off += TCG_TARGET_STACK_BIAS;
#endif
- ts->mem_base = s->frame_temp;
- ts->mem_allocated = 1;
+
+ /* If the object was subdivided, assign memory to all the parts. */
+ if (ts->base_type != ts->type) {
+ int part_size = tcg_type_size(ts->type);
+ int part_count = size / part_size;
+
+ ts -= ts->temp_subindex;
+ for (int i = 0; i < part_count; ++i) {
+ ts[i].mem_offset = off + i * part_size;
+ ts[i].mem_base = s->frame_temp;
+ ts[i].mem_allocated = 1;
+ }
+ } else {
+ ts->mem_offset = off;
+ ts->mem_base = s->frame_temp;
+ ts->mem_allocated = 1;
+ }
}
static void temp_load(TCGContext *, TCGTemp *, TCGRegSet, TCGRegSet, TCGRegSet);
--
2.34.1
^ permalink raw reply related [flat|nested] 43+ messages in thread* [PATCH v2 26/36] tcg: Introduce tcg_out_addi_ptr
2022-10-21 7:15 [PATCH v2 00/36] tcg: Support for Int128 with helpers Richard Henderson
` (24 preceding siblings ...)
2022-10-21 7:15 ` [PATCH v2 25/36] tcg: Allocate objects contiguously in temp_allocate_frame Richard Henderson
@ 2022-10-21 7:15 ` Richard Henderson
2022-10-21 7:15 ` [PATCH v2 27/36] tcg: Add TCG_CALL_{RET,ARG}_BY_REF Richard Henderson
` (9 subsequent siblings)
35 siblings, 0 replies; 43+ messages in thread
From: Richard Henderson @ 2022-10-21 7:15 UTC (permalink / raw)
To: qemu-devel
Implement the function for arm, i386, and s390x, which will use it.
Add stubs for all other backends.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
tcg/tcg.c | 2 ++
tcg/aarch64/tcg-target.c.inc | 7 +++++++
tcg/arm/tcg-target.c.inc | 20 ++++++++++++++++++++
tcg/i386/tcg-target.c.inc | 8 ++++++++
tcg/loongarch64/tcg-target.c.inc | 7 +++++++
tcg/mips/tcg-target.c.inc | 7 +++++++
tcg/ppc/tcg-target.c.inc | 7 +++++++
tcg/riscv/tcg-target.c.inc | 7 +++++++
tcg/s390x/tcg-target.c.inc | 7 +++++++
tcg/sparc64/tcg-target.c.inc | 7 +++++++
tcg/tci/tcg-target.c.inc | 7 +++++++
11 files changed, 86 insertions(+)
diff --git a/tcg/tcg.c b/tcg/tcg.c
index a3538edffa..10d239d15c 100644
--- a/tcg/tcg.c
+++ b/tcg/tcg.c
@@ -103,6 +103,8 @@ static void tcg_out_ld(TCGContext *s, TCGType type, TCGReg ret, TCGReg arg1,
static bool tcg_out_mov(TCGContext *s, TCGType type, TCGReg ret, TCGReg arg);
static void tcg_out_movi(TCGContext *s, TCGType type,
TCGReg ret, tcg_target_long arg);
+static void tcg_out_addi_ptr(TCGContext *s, TCGReg, TCGReg, tcg_target_long)
+ __attribute__((unused));
static void tcg_out_op(TCGContext *s, TCGOpcode opc,
const TCGArg args[TCG_MAX_OP_ARGS],
const int const_args[TCG_MAX_OP_ARGS]);
diff --git a/tcg/aarch64/tcg-target.c.inc b/tcg/aarch64/tcg-target.c.inc
index e8184fe001..f089a74f0e 100644
--- a/tcg/aarch64/tcg-target.c.inc
+++ b/tcg/aarch64/tcg-target.c.inc
@@ -1102,6 +1102,13 @@ static void tcg_out_movi(TCGContext *s, TCGType type, TCGReg rd,
tcg_out_insn(s, 3305, LDR, 0, rd);
}
+static void tcg_out_addi_ptr(TCGContext *s, TCGReg rd, TCGReg rs,
+ tcg_target_long imm)
+{
+ /* This function is only used for passing structs by reference. */
+ g_assert_not_reached();
+}
+
/* Define something more legible for general use. */
#define tcg_out_ldst_r tcg_out_insn_3310
diff --git a/tcg/arm/tcg-target.c.inc b/tcg/arm/tcg-target.c.inc
index e82749a602..2950a29d49 100644
--- a/tcg/arm/tcg-target.c.inc
+++ b/tcg/arm/tcg-target.c.inc
@@ -2548,6 +2548,26 @@ static void tcg_out_movi(TCGContext *s, TCGType type,
tcg_out_movi32(s, COND_AL, ret, arg);
}
+static void tcg_out_addi_ptr(TCGContext *s, TCGReg rd, TCGReg rs,
+ tcg_target_long imm)
+{
+ int enc, opc = ARITH_ADD;
+
+ /* All of the easiest immediates to encode are positive. */
+ if (imm < 0) {
+ imm = -imm;
+ opc = ARITH_SUB;
+ }
+ enc = encode_imm(imm);
+ if (enc >= 0) {
+ tcg_out_dat_imm(s, COND_AL, opc, rd, rs, enc);
+ } else {
+ tcg_out_movi32(s, COND_AL, TCG_REG_TMP, imm);
+ tcg_out_dat_reg(s, COND_AL, opc, rd, rs,
+ TCG_REG_TMP, SHIFT_IMM_LSL(0));
+ }
+}
+
/* Type is always V128, with I64 elements. */
static void tcg_out_dup2_vec(TCGContext *s, TCGReg rd, TCGReg rl, TCGReg rh)
{
diff --git a/tcg/i386/tcg-target.c.inc b/tcg/i386/tcg-target.c.inc
index 58bd5873f5..6a021dda8b 100644
--- a/tcg/i386/tcg-target.c.inc
+++ b/tcg/i386/tcg-target.c.inc
@@ -1069,6 +1069,14 @@ static void tcg_out_movi(TCGContext *s, TCGType type,
}
}
+static void tcg_out_addi_ptr(TCGContext *s, TCGReg rd, TCGReg rs,
+ tcg_target_long imm)
+{
+ /* This function is only used for passing structs by reference. */
+ tcg_debug_assert(TCG_TARGET_REG_BITS == 32);
+ tcg_out_modrm_offset(s, OPC_LEA, rd, rs, imm);
+}
+
static inline void tcg_out_pushi(TCGContext *s, tcg_target_long val)
{
if (val == (int8_t)val) {
diff --git a/tcg/loongarch64/tcg-target.c.inc b/tcg/loongarch64/tcg-target.c.inc
index 98da264f16..967bf307b8 100644
--- a/tcg/loongarch64/tcg-target.c.inc
+++ b/tcg/loongarch64/tcg-target.c.inc
@@ -389,6 +389,13 @@ static void tcg_out_movi(TCGContext *s, TCGType type, TCGReg rd,
}
}
+static void tcg_out_addi_ptr(TCGContext *s, TCGReg rd, TCGReg rs,
+ tcg_target_long imm)
+{
+ /* This function is only used for passing structs by reference. */
+ g_assert_not_reached();
+}
+
static void tcg_out_ext8u(TCGContext *s, TCGReg ret, TCGReg arg)
{
tcg_out_opc_andi(s, ret, arg, 0xff);
diff --git a/tcg/mips/tcg-target.c.inc b/tcg/mips/tcg-target.c.inc
index 292e490b5c..22b5463f0f 100644
--- a/tcg/mips/tcg-target.c.inc
+++ b/tcg/mips/tcg-target.c.inc
@@ -550,6 +550,13 @@ static void tcg_out_movi(TCGContext *s, TCGType type,
}
}
+static void tcg_out_addi_ptr(TCGContext *s, TCGReg rd, TCGReg rs,
+ tcg_target_long imm)
+{
+ /* This function is only used for passing structs by reference. */
+ g_assert_not_reached();
+}
+
static void tcg_out_bswap16(TCGContext *s, TCGReg ret, TCGReg arg, int flags)
{
/* ret and arg can't be register tmp0 */
diff --git a/tcg/ppc/tcg-target.c.inc b/tcg/ppc/tcg-target.c.inc
index f561a3492f..d9e4ba8883 100644
--- a/tcg/ppc/tcg-target.c.inc
+++ b/tcg/ppc/tcg-target.c.inc
@@ -1125,6 +1125,13 @@ static void tcg_out_movi(TCGContext *s, TCGType type, TCGReg ret,
}
}
+static void tcg_out_addi_ptr(TCGContext *s, TCGReg rd, TCGReg rs,
+ tcg_target_long imm)
+{
+ /* This function is only used for passing structs by reference. */
+ g_assert_not_reached();
+}
+
static bool mask_operand(uint32_t c, int *mb, int *me)
{
uint32_t lsb, test;
diff --git a/tcg/riscv/tcg-target.c.inc b/tcg/riscv/tcg-target.c.inc
index aa017d665a..6072945ccb 100644
--- a/tcg/riscv/tcg-target.c.inc
+++ b/tcg/riscv/tcg-target.c.inc
@@ -545,6 +545,13 @@ static void tcg_out_movi(TCGContext *s, TCGType type, TCGReg rd,
tcg_out_opc_imm(s, OPC_LD, rd, rd, 0);
}
+static void tcg_out_addi_ptr(TCGContext *s, TCGReg rd, TCGReg rs,
+ tcg_target_long imm)
+{
+ /* This function is only used for passing structs by reference. */
+ g_assert_not_reached();
+}
+
static void tcg_out_ext8u(TCGContext *s, TCGReg ret, TCGReg arg)
{
tcg_out_opc_imm(s, OPC_ANDI, ret, arg, 0xff);
diff --git a/tcg/s390x/tcg-target.c.inc b/tcg/s390x/tcg-target.c.inc
index c3c0bcc3eb..8663a963a6 100644
--- a/tcg/s390x/tcg-target.c.inc
+++ b/tcg/s390x/tcg-target.c.inc
@@ -1020,6 +1020,13 @@ static inline bool tcg_out_sti(TCGContext *s, TCGType type, TCGArg val,
return false;
}
+static void tcg_out_addi_ptr(TCGContext *s, TCGReg rd, TCGReg rs,
+ tcg_target_long imm)
+{
+ /* This function is only used for passing structs by reference. */
+ tcg_out_mem(s, RX_LA, RXY_LAY, rd, rs, TCG_REG_NONE, imm);
+}
+
/* load data from an absolute host address */
static void tcg_out_ld_abs(TCGContext *s, TCGType type,
TCGReg dest, const void *abs)
diff --git a/tcg/sparc64/tcg-target.c.inc b/tcg/sparc64/tcg-target.c.inc
index eb913f33c8..f6a8a8e605 100644
--- a/tcg/sparc64/tcg-target.c.inc
+++ b/tcg/sparc64/tcg-target.c.inc
@@ -497,6 +497,13 @@ static void tcg_out_movi(TCGContext *s, TCGType type,
tcg_out_movi_int(s, type, ret, arg, false, TCG_REG_T2);
}
+static void tcg_out_addi_ptr(TCGContext *s, TCGReg rd, TCGReg rs,
+ tcg_target_long imm)
+{
+ /* This function is only used for passing structs by reference. */
+ g_assert_not_reached();
+}
+
static void tcg_out_ldst_rr(TCGContext *s, TCGReg data, TCGReg a1,
TCGReg a2, int op)
{
diff --git a/tcg/tci/tcg-target.c.inc b/tcg/tci/tcg-target.c.inc
index b07b30b9fc..8bf02a96e9 100644
--- a/tcg/tci/tcg-target.c.inc
+++ b/tcg/tci/tcg-target.c.inc
@@ -561,6 +561,13 @@ static void tcg_out_movi(TCGContext *s, TCGType type,
}
}
+static void tcg_out_addi_ptr(TCGContext *s, TCGReg rd, TCGReg rs,
+ tcg_target_long imm)
+{
+ /* This function is only used for passing structs by reference. */
+ g_assert_not_reached();
+}
+
static void tcg_out_call(TCGContext *s, const tcg_insn_unit *func,
const TCGHelperInfo *info)
{
--
2.34.1
^ permalink raw reply related [flat|nested] 43+ messages in thread* [PATCH v2 27/36] tcg: Add TCG_CALL_{RET,ARG}_BY_REF
2022-10-21 7:15 [PATCH v2 00/36] tcg: Support for Int128 with helpers Richard Henderson
` (25 preceding siblings ...)
2022-10-21 7:15 ` [PATCH v2 26/36] tcg: Introduce tcg_out_addi_ptr Richard Henderson
@ 2022-10-21 7:15 ` Richard Henderson
2022-10-21 7:15 ` [PATCH v2 28/36] tcg: Introduce tcg_target_call_oarg_reg Richard Henderson
` (8 subsequent siblings)
35 siblings, 0 replies; 43+ messages in thread
From: Richard Henderson @ 2022-10-21 7:15 UTC (permalink / raw)
To: qemu-devel
These will be used by some hosts, 32 and 64-bit to pass and
return i128. Not yet used, because allocation is not yet enabled.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
tcg/tcg-internal.h | 10 ++--
tcg/tcg.c | 124 ++++++++++++++++++++++++++++++++++++++++++++-
2 files changed, 129 insertions(+), 5 deletions(-)
diff --git a/tcg/tcg-internal.h b/tcg/tcg-internal.h
index 7d4f7049be..1fe7bd7d5d 100644
--- a/tcg/tcg-internal.h
+++ b/tcg/tcg-internal.h
@@ -37,6 +37,7 @@
typedef enum {
TCG_CALL_RET_NORMAL, /* by registers */
TCG_CALL_RET_NORMAL_4, /* for i128, by 4 registers */
+ TCG_CALL_RET_BY_REF, /* for i128, by reference as first arg */
} TCGCallReturnKind;
typedef enum {
@@ -46,12 +47,15 @@ typedef enum {
TCG_CALL_ARG_EXTEND_U, /* ... as a zero-extended i64 */
TCG_CALL_ARG_EXTEND_S, /* ... as a sign-extended i64 */
TCG_CALL_ARG_NORMAL_4, /* for i128, like normal with 4 slots */
+ TCG_CALL_ARG_BY_REF, /* ... by reference, first */
+ TCG_CALL_ARG_BY_REF_2, /* ... by reference, second */
} TCGCallArgumentKind;
typedef struct TCGCallArgumentLoc {
- TCGCallArgumentKind kind : 8;
- unsigned reg_slot : 8;
- unsigned stk_slot : 8;
+ TCGCallArgumentKind kind : 4;
+ unsigned reg_slot : 6;
+ unsigned stk_slot : 6;
+ unsigned ref_slot : 6;
unsigned reg_n : 2;
unsigned arg_idx : 4;
unsigned tmp_subindex : 1;
diff --git a/tcg/tcg.c b/tcg/tcg.c
index 10d239d15c..0e141807c6 100644
--- a/tcg/tcg.c
+++ b/tcg/tcg.c
@@ -103,8 +103,7 @@ static void tcg_out_ld(TCGContext *s, TCGType type, TCGReg ret, TCGReg arg1,
static bool tcg_out_mov(TCGContext *s, TCGType type, TCGReg ret, TCGReg arg);
static void tcg_out_movi(TCGContext *s, TCGType type,
TCGReg ret, tcg_target_long arg);
-static void tcg_out_addi_ptr(TCGContext *s, TCGReg, TCGReg, tcg_target_long)
- __attribute__((unused));
+static void tcg_out_addi_ptr(TCGContext *s, TCGReg, TCGReg, tcg_target_long);
static void tcg_out_op(TCGContext *s, TCGOpcode opc,
const TCGArg args[TCG_MAX_OP_ARGS],
const int const_args[TCG_MAX_OP_ARGS]);
@@ -684,6 +683,34 @@ static void layout_arg_normal_4(TCGCumulativeArgs *cum, TCGHelperInfo *info)
cum->op_arg_idx++;
}
+static void layout_arg_by_ref(TCGCumulativeArgs *cum, TCGHelperInfo *info)
+{
+ TCGCallArgumentLoc *loc = &info->in[cum->info_in_idx];
+ int n = 128 / TCG_TARGET_REG_BITS;
+
+ /* The first subindex carries the pointer. */
+ layout_arg_1(cum, info, TCG_CALL_ARG_BY_REF);
+
+ /*
+ * The callee is allowed to clobber memory associated with
+ * structure pass by-reference. Therefore we must make copies.
+ * Allocate space from the top of TCG_STATIC_CALL_ARGS_SIZE.
+ */
+ cum->max_stk_slot -= n;
+ loc[0].ref_slot = cum->max_stk_slot;
+
+ /* Any other subindex is present for liveness. */
+ if (TCG_TARGET_REG_BITS == 64) {
+ loc[1] = (TCGCallArgumentLoc){
+ .kind = TCG_CALL_ARG_BY_REF_2,
+ .arg_idx = cum->arg_idx,
+ .tmp_subindex = 1,
+ };
+ cum->info_in_idx++;
+ cum->op_arg_idx++;
+ }
+}
+
static void init_call_layout(TCGHelperInfo *info)
{
unsigned typemask = info->typemask;
@@ -728,6 +755,18 @@ static void init_call_layout(TCGHelperInfo *info)
assert(ARRAY_SIZE(tcg_target_call_oarg_regs) >= 2);
}
break;
+ case TCG_CALL_RET_BY_REF:
+ /*
+ * Allocate the first argument to the output.
+ * We don't need to store this anywhere, just make it
+ * unavailable for use in the input loop below.
+ */
+ if (cum.max_reg_slot == 0) {
+ cum.stk_slot = 1;
+ } else {
+ cum.reg_slot = 1;
+ }
+ break;
default:
g_assert_not_reached();
}
@@ -765,6 +804,10 @@ static void init_call_layout(TCGHelperInfo *info)
type = TCG_TYPE_PTR;
kind = TCG_CALL_ARG_NORMAL;
break;
+ case dh_typecode_i128:
+ type = TCG_TYPE_I128;
+ kind = TCG_CALL_ARG_NORMAL; /* TODO */
+ break;
default:
g_assert_not_reached();
}
@@ -809,6 +852,10 @@ static void init_call_layout(TCGHelperInfo *info)
assert(type == TCG_TYPE_I32);
layout_arg_1(&cum, info, kind);
break;
+ case TCG_CALL_ARG_BY_REF:
+ assert(type == TCG_TYPE_I128);
+ layout_arg_by_ref(&cum, info);
+ break;
default:
g_assert_not_reached();
}
@@ -1740,6 +1787,8 @@ void tcg_gen_callN(void *func, TCGTemp *ret, int nargs, TCGTemp **args)
switch (loc->kind) {
case TCG_CALL_ARG_NORMAL:
case TCG_CALL_ARG_NORMAL_4:
+ case TCG_CALL_ARG_BY_REF:
+ case TCG_CALL_ARG_BY_REF_2:
op->args[pi++] = temp_arg(ts);
break;
@@ -4466,6 +4515,27 @@ static void load_arg_normal_4(TCGContext *s, const TCGCallArgumentLoc *loc,
}
}
+static void load_arg_by_ref(TCGContext *s, const TCGCallArgumentLoc *loc,
+ TCGReg ref_base, int ref_off,
+ TCGRegSet *allocated_regs)
+{
+ TCGReg reg;
+
+ if (loc->reg_n) {
+ reg = tcg_target_call_iarg_regs[loc->reg_slot];
+ tcg_out_addi_ptr(s, reg, ref_base, ref_off);
+ tcg_regset_set_reg(*allocated_regs, reg);
+ } else {
+ int stk_off = TCG_TARGET_CALL_STACK_OFFSET
+ + loc->stk_slot * sizeof(tcg_target_long);
+
+ reg = tcg_reg_alloc(s, tcg_target_available_regs[TCG_TYPE_PTR],
+ *allocated_regs, 0, false);
+ tcg_out_addi_ptr(s, reg, ref_base, ref_off);
+ tcg_out_st(s, TCG_TYPE_PTR, reg, TCG_REG_CALL_STACK, stk_off);
+ }
+}
+
static void tcg_reg_alloc_call(TCGContext *s, TCGOp *op)
{
const int nb_oargs = TCGOP_CALLO(op);
@@ -4475,6 +4545,19 @@ static void tcg_reg_alloc_call(TCGContext *s, TCGOp *op)
TCGRegSet allocated_regs = s->reserved_regs;
int i;
+ /*
+ * The callee is allowed to clobber memory associated with
+ * structure pass by-reference. Therefore we must make copies.
+ * Do this first, before we start allocating argument regs.
+ */
+ for (i = 0; i < nb_iargs; ++i) {
+ const TCGCallArgumentLoc *loc = &info->in[i];
+ if (loc->kind == TCG_CALL_ARG_BY_REF) {
+ TCGTemp *ts = arg_temp(op->args[nb_oargs + i]);
+ copy_to_stk_i128(s, loc->ref_slot, ts, 0, allocated_regs);
+ }
+ }
+
/*
* Move inputs into place in reverse order,
* so that we place stacked arguments first.
@@ -4492,6 +4575,15 @@ static void tcg_reg_alloc_call(TCGContext *s, TCGOp *op)
case TCG_CALL_ARG_NORMAL_4:
load_arg_normal_4(s, loc, ts, &allocated_regs);
break;
+ case TCG_CALL_ARG_BY_REF:
+ load_arg_by_ref(s, loc, TCG_REG_CALL_STACK,
+ TCG_TARGET_CALL_STACK_OFFSET +
+ loc->ref_slot * sizeof(tcg_target_long),
+ &allocated_regs);
+ break;
+ case TCG_CALL_ARG_BY_REF_2:
+ /* nothing to do */
+ break;
default:
g_assert_not_reached();
}
@@ -4523,6 +4615,26 @@ static void tcg_reg_alloc_call(TCGContext *s, TCGOp *op)
save_globals(s, allocated_regs);
}
+ /*
+ * If the ABI passes a pointer to the returned struct as the first
+ * argument, load that now. Pass a pointer to the output home slot.
+ */
+ if (info->out_kind == TCG_CALL_RET_BY_REF) {
+ static const TCGCallArgumentLoc ret_loc = {
+ .reg_n = ARRAY_SIZE(tcg_target_call_iarg_regs) != 0
+ };
+ TCGTemp *ts = arg_temp(op->args[0]);
+
+ tcg_debug_assert(ts->base_type == TCG_TYPE_I128);
+ tcg_debug_assert(ts->temp_subindex == 0);
+
+ if (!ts->mem_allocated) {
+ temp_allocate_frame(s, ts);
+ }
+ load_arg_by_ref(s, &ret_loc, ts->mem_base->reg, ts->mem_offset,
+ &allocated_regs);
+ }
+
tcg_out_call(s, tcg_call_func(op), info);
/* Assign output registers and emit moves if needed. */
@@ -4563,6 +4675,14 @@ static void tcg_reg_alloc_call(TCGContext *s, TCGOp *op)
}
break;
+ case TCG_CALL_RET_BY_REF:
+ /* The callee has performed a write through the reference. */
+ for (i = 0; i < nb_oargs; i++) {
+ TCGTemp *ts = arg_temp(op->args[i]);
+ ts->val_type = TEMP_VAL_MEM;
+ }
+ break;
+
default:
g_assert_not_reached();
}
--
2.34.1
^ permalink raw reply related [flat|nested] 43+ messages in thread* [PATCH v2 28/36] tcg: Introduce tcg_target_call_oarg_reg
2022-10-21 7:15 [PATCH v2 00/36] tcg: Support for Int128 with helpers Richard Henderson
` (26 preceding siblings ...)
2022-10-21 7:15 ` [PATCH v2 27/36] tcg: Add TCG_CALL_{RET,ARG}_BY_REF Richard Henderson
@ 2022-10-21 7:15 ` Richard Henderson
2022-10-21 7:15 ` [PATCH v2 29/36] tcg: Add TCG_CALL_RET_BY_VEC Richard Henderson
` (7 subsequent siblings)
35 siblings, 0 replies; 43+ messages in thread
From: Richard Henderson @ 2022-10-21 7:15 UTC (permalink / raw)
To: qemu-devel
Replace the flat array tcg_target_call_oarg_regs[] with
a function call including the TCGCallReturnKind.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
tcg/tcg.c | 14 ++++++++------
tcg/aarch64/tcg-target.c.inc | 10 +++++++---
tcg/arm/tcg-target.c.inc | 9 ++++++---
tcg/i386/tcg-target.c.inc | 22 ++++++++++++++++------
tcg/loongarch64/tcg-target.c.inc | 10 ++++++----
tcg/mips/tcg-target.c.inc | 10 ++++++----
tcg/ppc/tcg-target.c.inc | 10 ++++++----
tcg/riscv/tcg-target.c.inc | 10 ++++++----
tcg/s390x/tcg-target.c.inc | 9 ++++++---
tcg/sparc64/tcg-target.c.inc | 12 ++++++------
tcg/tci/tcg-target.c.inc | 11 +++++------
11 files changed, 78 insertions(+), 49 deletions(-)
diff --git a/tcg/tcg.c b/tcg/tcg.c
index 0e141807c6..25dc3c9f8f 100644
--- a/tcg/tcg.c
+++ b/tcg/tcg.c
@@ -148,6 +148,7 @@ static bool tcg_out_sti(TCGContext *s, TCGType type, TCGArg val,
TCGReg base, intptr_t ofs);
static void tcg_out_call(TCGContext *s, const tcg_insn_unit *target,
const TCGHelperInfo *info);
+static TCGReg tcg_target_call_oarg_reg(TCGCallReturnKind kind, int slot);
static bool tcg_target_const_match(int64_t val, TCGType type, int ct);
#ifdef TCG_TARGET_NEED_LDST_LABELS
static int tcg_out_ldst_finalize(TCGContext *s);
@@ -749,11 +750,11 @@ static void init_call_layout(TCGHelperInfo *info)
switch (/* TODO */ TCG_CALL_RET_NORMAL) {
case TCG_CALL_RET_NORMAL:
if (TCG_TARGET_REG_BITS == 32) {
- assert(ARRAY_SIZE(tcg_target_call_oarg_regs) >= 4);
info->out_kind = TCG_CALL_RET_NORMAL_4;
- } else {
- assert(ARRAY_SIZE(tcg_target_call_oarg_regs) >= 2);
}
+ /* Query the register now to trigger any assert early. */
+ (void)tcg_target_call_oarg_reg(info->out_kind,
+ 127 / TCG_TARGET_REG_BITS);
break;
case TCG_CALL_RET_BY_REF:
/*
@@ -2826,7 +2827,7 @@ static void liveness_pass_1(TCGContext *s)
ts->state = TS_DEAD;
la_reset_pref(ts);
- /* Not used -- it will be tcg_target_call_oarg_regs[i]. */
+ /* Not used -- it will be tcg_target_call_oarg_reg(). */
op->output_pref[i] = 0;
}
@@ -4642,7 +4643,7 @@ static void tcg_reg_alloc_call(TCGContext *s, TCGOp *op)
case TCG_CALL_RET_NORMAL:
for (i = 0; i < nb_oargs; i++) {
TCGTemp *ts = arg_temp(op->args[i]);
- TCGReg reg = tcg_target_call_oarg_regs[i];
+ TCGReg reg = tcg_target_call_oarg_reg(TCG_CALL_RET_NORMAL, i);
/* ENV should not be modified. */
tcg_debug_assert(!temp_readonly(ts));
@@ -4668,7 +4669,8 @@ static void tcg_reg_alloc_call(TCGContext *s, TCGOp *op)
temp_allocate_frame(s, ts);
}
for (i = 0; i < 4; i++) {
- tcg_out_st(s, TCG_TYPE_I32, tcg_target_call_oarg_regs[i],
+ tcg_out_st(s, TCG_TYPE_I32,
+ tcg_target_call_oarg_reg(TCG_CALL_RET_NORMAL_4, i),
ts->mem_base->reg, ts->mem_offset + i * 4);
}
ts->val_type = TEMP_VAL_MEM;
diff --git a/tcg/aarch64/tcg-target.c.inc b/tcg/aarch64/tcg-target.c.inc
index f089a74f0e..dc99fa3257 100644
--- a/tcg/aarch64/tcg-target.c.inc
+++ b/tcg/aarch64/tcg-target.c.inc
@@ -63,9 +63,13 @@ static const int tcg_target_call_iarg_regs[8] = {
TCG_REG_X0, TCG_REG_X1, TCG_REG_X2, TCG_REG_X3,
TCG_REG_X4, TCG_REG_X5, TCG_REG_X6, TCG_REG_X7
};
-static const int tcg_target_call_oarg_regs[1] = {
- TCG_REG_X0
-};
+
+static TCGReg tcg_target_call_oarg_reg(TCGCallReturnKind kind, int slot)
+{
+ tcg_debug_assert(kind == TCG_CALL_RET_NORMAL);
+ tcg_debug_assert(slot >= 0 && slot <= 1);
+ return TCG_REG_X0 + slot;
+}
#define TCG_REG_TMP TCG_REG_X30
#define TCG_VEC_TMP TCG_REG_V31
diff --git a/tcg/arm/tcg-target.c.inc b/tcg/arm/tcg-target.c.inc
index 2950a29d49..41d3af517d 100644
--- a/tcg/arm/tcg-target.c.inc
+++ b/tcg/arm/tcg-target.c.inc
@@ -79,9 +79,12 @@ static const int tcg_target_reg_alloc_order[] = {
static const int tcg_target_call_iarg_regs[4] = {
TCG_REG_R0, TCG_REG_R1, TCG_REG_R2, TCG_REG_R3
};
-static const int tcg_target_call_oarg_regs[2] = {
- TCG_REG_R0, TCG_REG_R1
-};
+
+static TCGReg tcg_target_call_oarg_reg(TCGCallReturnKind kind, int slot)
+{
+ tcg_debug_assert(slot >= 0 && slot <= 3);
+ return TCG_REG_R0 + slot;
+}
#define TCG_REG_TMP TCG_REG_R12
#define TCG_VEC_TMP TCG_REG_Q15
diff --git a/tcg/i386/tcg-target.c.inc b/tcg/i386/tcg-target.c.inc
index 6a021dda8b..82c8491152 100644
--- a/tcg/i386/tcg-target.c.inc
+++ b/tcg/i386/tcg-target.c.inc
@@ -109,12 +109,22 @@ static const int tcg_target_call_iarg_regs[] = {
#endif
};
-static const int tcg_target_call_oarg_regs[] = {
- TCG_REG_EAX,
-#if TCG_TARGET_REG_BITS == 32
- TCG_REG_EDX
-#endif
-};
+static TCGReg tcg_target_call_oarg_reg(TCGCallReturnKind kind, int slot)
+{
+ switch (kind) {
+ case TCG_CALL_RET_NORMAL:
+ switch (slot) {
+ case 0:
+ return TCG_REG_EAX;
+ case 1:
+ return TCG_REG_EDX;
+ }
+ break;
+ default:
+ break;
+ }
+ g_assert_not_reached();
+}
/* Constants we accept. */
#define TCG_CT_CONST_S32 0x100
diff --git a/tcg/loongarch64/tcg-target.c.inc b/tcg/loongarch64/tcg-target.c.inc
index 967bf307b8..2efefe863e 100644
--- a/tcg/loongarch64/tcg-target.c.inc
+++ b/tcg/loongarch64/tcg-target.c.inc
@@ -114,10 +114,12 @@ static const int tcg_target_call_iarg_regs[] = {
TCG_REG_A7,
};
-static const int tcg_target_call_oarg_regs[] = {
- TCG_REG_A0,
- TCG_REG_A1,
-};
+static TCGReg tcg_target_call_oarg_reg(TCGCallReturnKind kind, int slot)
+{
+ tcg_debug_assert(kind == TCG_CALL_RET_NORMAL);
+ tcg_debug_assert(slot >= 0 && slot <= 1);
+ return TCG_REG_A0 + slot;
+}
#ifndef CONFIG_SOFTMMU
#define USE_GUEST_BASE (guest_base != 0)
diff --git a/tcg/mips/tcg-target.c.inc b/tcg/mips/tcg-target.c.inc
index 22b5463f0f..92883176c6 100644
--- a/tcg/mips/tcg-target.c.inc
+++ b/tcg/mips/tcg-target.c.inc
@@ -136,10 +136,12 @@ static const TCGReg tcg_target_call_iarg_regs[] = {
#endif
};
-static const TCGReg tcg_target_call_oarg_regs[2] = {
- TCG_REG_V0,
- TCG_REG_V1
-};
+static TCGReg tcg_target_call_oarg_reg(TCGCallReturnKind kind, int slot)
+{
+ tcg_debug_assert(kind == TCG_CALL_RET_NORMAL);
+ tcg_debug_assert(slot >= 0 && slot <= 1);
+ return TCG_REG_V0 + slot;
+}
static const tcg_insn_unit *tb_ret_addr;
static const tcg_insn_unit *bswap32_addr;
diff --git a/tcg/ppc/tcg-target.c.inc b/tcg/ppc/tcg-target.c.inc
index d9e4ba8883..781ecfe161 100644
--- a/tcg/ppc/tcg-target.c.inc
+++ b/tcg/ppc/tcg-target.c.inc
@@ -186,10 +186,12 @@ static const int tcg_target_call_iarg_regs[] = {
TCG_REG_R10
};
-static const int tcg_target_call_oarg_regs[] = {
- TCG_REG_R3,
- TCG_REG_R4
-};
+static TCGReg tcg_target_call_oarg_reg(TCGCallReturnKind kind, int slot)
+{
+ tcg_debug_assert(kind == TCG_CALL_RET_NORMAL);
+ tcg_debug_assert(slot >= 0 && slot <= 1);
+ return TCG_REG_R3 + slot;
+}
static const int tcg_target_callee_save_regs[] = {
#ifdef _CALL_DARWIN
diff --git a/tcg/riscv/tcg-target.c.inc b/tcg/riscv/tcg-target.c.inc
index 6072945ccb..417736cae7 100644
--- a/tcg/riscv/tcg-target.c.inc
+++ b/tcg/riscv/tcg-target.c.inc
@@ -113,10 +113,12 @@ static const int tcg_target_call_iarg_regs[] = {
TCG_REG_A7,
};
-static const int tcg_target_call_oarg_regs[] = {
- TCG_REG_A0,
- TCG_REG_A1,
-};
+static TCGReg tcg_target_call_oarg_reg(TCGCallReturnKind kind, int slot)
+{
+ tcg_debug_assert(kind == TCG_CALL_RET_NORMAL);
+ tcg_debug_assert(slot >= 0 && slot <= 1);
+ return TCG_REG_A0 + slot;
+}
#define TCG_CT_CONST_ZERO 0x100
#define TCG_CT_CONST_S12 0x200
diff --git a/tcg/s390x/tcg-target.c.inc b/tcg/s390x/tcg-target.c.inc
index 8663a963a6..50655e9d1d 100644
--- a/tcg/s390x/tcg-target.c.inc
+++ b/tcg/s390x/tcg-target.c.inc
@@ -390,9 +390,12 @@ static const int tcg_target_call_iarg_regs[] = {
TCG_REG_R6,
};
-static const int tcg_target_call_oarg_regs[] = {
- TCG_REG_R2,
-};
+static TCGReg tcg_target_call_oarg_reg(TCGCallReturnKind kind, int slot)
+{
+ tcg_debug_assert(kind == TCG_CALL_RET_NORMAL);
+ tcg_debug_assert(slot == 0);
+ return TCG_REG_R2;
+}
#define S390_CC_EQ 8
#define S390_CC_LT 4
diff --git a/tcg/sparc64/tcg-target.c.inc b/tcg/sparc64/tcg-target.c.inc
index f6a8a8e605..9b5afb8248 100644
--- a/tcg/sparc64/tcg-target.c.inc
+++ b/tcg/sparc64/tcg-target.c.inc
@@ -133,12 +133,12 @@ static const int tcg_target_call_iarg_regs[6] = {
TCG_REG_O5,
};
-static const int tcg_target_call_oarg_regs[] = {
- TCG_REG_O0,
- TCG_REG_O1,
- TCG_REG_O2,
- TCG_REG_O3,
-};
+static TCGReg tcg_target_call_oarg_reg(TCGCallReturnKind kind, int slot)
+{
+ tcg_debug_assert(kind == TCG_CALL_RET_NORMAL);
+ tcg_debug_assert(slot >= 0 && slot <= 3);
+ return TCG_REG_O0 + slot;
+}
#define INSN_OP(x) ((x) << 30)
#define INSN_OP2(x) ((x) << 22)
diff --git a/tcg/tci/tcg-target.c.inc b/tcg/tci/tcg-target.c.inc
index 8bf02a96e9..914806c216 100644
--- a/tcg/tci/tcg-target.c.inc
+++ b/tcg/tci/tcg-target.c.inc
@@ -204,12 +204,11 @@ static const int tcg_target_reg_alloc_order[] = {
/* No call arguments via registers. All will be stored on the "stack". */
static const int tcg_target_call_iarg_regs[] = { };
-static const int tcg_target_call_oarg_regs[] = {
- TCG_REG_R0,
-#if TCG_TARGET_REG_BITS == 32
- TCG_REG_R1
-#endif
-};
+static TCGReg tcg_target_call_oarg_reg(TCGCallReturnKind kind, int slot)
+{
+ tcg_debug_assert(slot >= 0 && slot <= 3);
+ return TCG_REG_R0 + slot;
+}
#ifdef CONFIG_DEBUG_TCG
static const char *const tcg_target_reg_names[TCG_TARGET_NB_REGS] = {
--
2.34.1
^ permalink raw reply related [flat|nested] 43+ messages in thread* [PATCH v2 29/36] tcg: Add TCG_CALL_RET_BY_VEC
2022-10-21 7:15 [PATCH v2 00/36] tcg: Support for Int128 with helpers Richard Henderson
` (27 preceding siblings ...)
2022-10-21 7:15 ` [PATCH v2 28/36] tcg: Introduce tcg_target_call_oarg_reg Richard Henderson
@ 2022-10-21 7:15 ` Richard Henderson
2022-10-21 7:15 ` [PATCH v2 30/36] include/qemu/int128: Use Int128 structure for TCI Richard Henderson
` (6 subsequent siblings)
35 siblings, 0 replies; 43+ messages in thread
From: Richard Henderson @ 2022-10-21 7:15 UTC (permalink / raw)
To: qemu-devel
This will be used by _WIN64 to return i128. Not yet used,
because allocation is not yet enabled.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
tcg/tcg-internal.h | 1 +
tcg/tcg.c | 17 +++++++++++++++++
2 files changed, 18 insertions(+)
diff --git a/tcg/tcg-internal.h b/tcg/tcg-internal.h
index 1fe7bd7d5d..44ef51ca30 100644
--- a/tcg/tcg-internal.h
+++ b/tcg/tcg-internal.h
@@ -38,6 +38,7 @@ typedef enum {
TCG_CALL_RET_NORMAL, /* by registers */
TCG_CALL_RET_NORMAL_4, /* for i128, by 4 registers */
TCG_CALL_RET_BY_REF, /* for i128, by reference as first arg */
+ TCG_CALL_RET_BY_VEC, /* for i128, by vector register */
} TCGCallReturnKind;
typedef enum {
diff --git a/tcg/tcg.c b/tcg/tcg.c
index 25dc3c9f8f..9ec5b85f44 100644
--- a/tcg/tcg.c
+++ b/tcg/tcg.c
@@ -768,6 +768,8 @@ static void init_call_layout(TCGHelperInfo *info)
cum.reg_slot = 1;
}
break;
+ case TCG_CALL_RET_BY_VEC:
+ break;
default:
g_assert_not_reached();
}
@@ -4685,6 +4687,21 @@ static void tcg_reg_alloc_call(TCGContext *s, TCGOp *op)
}
break;
+ case TCG_CALL_RET_BY_VEC:
+ {
+ TCGTemp *ts = arg_temp(op->args[0]);
+
+ tcg_debug_assert(ts->type == TCG_TYPE_I128);
+ if (!ts->mem_allocated) {
+ temp_allocate_frame(s, ts);
+ }
+ tcg_out_st(s, TCG_TYPE_V128,
+ tcg_target_call_oarg_reg(TCG_CALL_RET_BY_VEC, 0),
+ ts->mem_base->reg, ts->mem_offset + i * 4);
+ ts->val_type = TEMP_VAL_MEM;
+ }
+ break;
+
default:
g_assert_not_reached();
}
--
2.34.1
^ permalink raw reply related [flat|nested] 43+ messages in thread* [PATCH v2 30/36] include/qemu/int128: Use Int128 structure for TCI
2022-10-21 7:15 [PATCH v2 00/36] tcg: Support for Int128 with helpers Richard Henderson
` (28 preceding siblings ...)
2022-10-21 7:15 ` [PATCH v2 29/36] tcg: Add TCG_CALL_RET_BY_VEC Richard Henderson
@ 2022-10-21 7:15 ` Richard Henderson
2022-10-21 7:15 ` [PATCH v2 31/36] tcg/i386: Add TCG_TARGET_CALL_{RET,ARG}_I128 Richard Henderson
` (5 subsequent siblings)
35 siblings, 0 replies; 43+ messages in thread
From: Richard Henderson @ 2022-10-21 7:15 UTC (permalink / raw)
To: qemu-devel
We are about to allow passing Int128 to/from tcg helper functions,
but libffi doesn't support __int128_t, so use the structure.
In order for atomic128.h to continue working, we must provide
a mechanism to frob between real __int128_t and the structure.
Provide a new union, Int128Alias, for this. We cannot modify
Int128 itself, as any changed alignment would also break libffi.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
include/qemu/atomic128.h | 34 ++++++++++++++++++++++++--------
include/qemu/int128.h | 18 ++++++++++++++---
util/int128.c | 42 ++++++++++++++++++++++++++++++++++++++++
3 files changed, 83 insertions(+), 11 deletions(-)
diff --git a/include/qemu/atomic128.h b/include/qemu/atomic128.h
index d179c05ede..5ba53f2f10 100644
--- a/include/qemu/atomic128.h
+++ b/include/qemu/atomic128.h
@@ -44,13 +44,23 @@
#if defined(CONFIG_ATOMIC128)
static inline Int128 atomic16_cmpxchg(Int128 *ptr, Int128 cmp, Int128 new)
{
- return qatomic_cmpxchg__nocheck(ptr, cmp, new);
+ Int128Alias r, c, n;
+
+ c.s = cmp;
+ n.s = new;
+ r.i = qatomic_cmpxchg__nocheck((__int128_t *)ptr, c.i, n.i);
+ return r.s;
}
# define HAVE_CMPXCHG128 1
#elif defined(CONFIG_CMPXCHG128)
static inline Int128 atomic16_cmpxchg(Int128 *ptr, Int128 cmp, Int128 new)
{
- return __sync_val_compare_and_swap_16(ptr, cmp, new);
+ Int128Alias r, c, n;
+
+ c.s = cmp;
+ n.s = new;
+ r.i = __sync_val_compare_and_swap_16((__int128_t *)ptr, c.i, n.i);
+ return r.s;
}
# define HAVE_CMPXCHG128 1
#elif defined(__aarch64__)
@@ -89,12 +99,18 @@ Int128 QEMU_ERROR("unsupported atomic")
#if defined(CONFIG_ATOMIC128)
static inline Int128 atomic16_read(Int128 *ptr)
{
- return qatomic_read__nocheck(ptr);
+ Int128Alias r;
+
+ r.i = qatomic_read__nocheck((__int128_t *)ptr);
+ return r.s;
}
static inline void atomic16_set(Int128 *ptr, Int128 val)
{
- qatomic_set__nocheck(ptr, val);
+ Int128Alias v;
+
+ v.s = val;
+ qatomic_set__nocheck((__int128_t *)ptr, v.i);
}
# define HAVE_ATOMIC128 1
@@ -153,7 +169,8 @@ static inline Int128 atomic16_read(Int128 *ptr)
if (have_atomic128) {
asm("vmovdqa %1, %0" : "=x" (ret) : "m" (*ptr));
} else {
- ret = atomic16_cmpxchg(ptr, 0, 0);
+ Int128 z = int128_make64(0);
+ ret = atomic16_cmpxchg(ptr, z, z);
}
return ret;
}
@@ -167,7 +184,7 @@ static inline void atomic16_set(Int128 *ptr, Int128 val)
do {
cmp = old;
old = atomic16_cmpxchg(ptr, cmp, val);
- } while (old != cmp);
+ } while (int128_ne(old, cmp));
}
}
@@ -176,7 +193,8 @@ static inline void atomic16_set(Int128 *ptr, Int128 val)
static inline Int128 atomic16_read(Int128 *ptr)
{
/* Maybe replace 0 with 0, returning the old value. */
- return atomic16_cmpxchg(ptr, 0, 0);
+ Int128 z = int128_make64(0);
+ return atomic16_cmpxchg(ptr, z, z);
}
static inline void atomic16_set(Int128 *ptr, Int128 val)
@@ -185,7 +203,7 @@ static inline void atomic16_set(Int128 *ptr, Int128 val)
do {
cmp = old;
old = atomic16_cmpxchg(ptr, cmp, val);
- } while (old != cmp);
+ } while (int128_ne(old, cmp));
}
# define HAVE_ATOMIC128 1
diff --git a/include/qemu/int128.h b/include/qemu/int128.h
index d2b76ca6ac..a062284025 100644
--- a/include/qemu/int128.h
+++ b/include/qemu/int128.h
@@ -3,7 +3,12 @@
#include "qemu/bswap.h"
-#ifdef CONFIG_INT128
+/*
+ * With TCI, we need to use libffi for interfacing with TCG helpers.
+ * But libffi does not support __int128_t, and therefore cannot pass
+ * or return values of this type, force use of the Int128 struct.
+ */
+#if defined(CONFIG_INT128) && !defined(CONFIG_TCG_INTERPRETER)
typedef __int128_t Int128;
static inline Int128 int128_make64(uint64_t a)
@@ -460,8 +465,7 @@ Int128 int128_divu(Int128, Int128);
Int128 int128_remu(Int128, Int128);
Int128 int128_divs(Int128, Int128);
Int128 int128_rems(Int128, Int128);
-
-#endif /* CONFIG_INT128 */
+#endif /* CONFIG_INT128 && !CONFIG_TCG_INTERPRETER */
static inline void bswap128s(Int128 *s)
{
@@ -472,4 +476,12 @@ static inline void bswap128s(Int128 *s)
#define INT128_MAX int128_make128(UINT64_MAX, INT64_MAX)
#define INT128_MIN int128_make128(0, INT64_MIN)
+#ifdef CONFIG_INT128
+typedef union {
+ Int128 s;
+ __int128_t i;
+ __uint128_t u;
+} Int128Alias;
+#endif /* CONFIG_INT128 */
+
#endif /* INT128_H */
diff --git a/util/int128.c b/util/int128.c
index ed8f25fef1..df6c6331bd 100644
--- a/util/int128.c
+++ b/util/int128.c
@@ -144,4 +144,46 @@ Int128 int128_rems(Int128 a, Int128 b)
return r;
}
+#elif defined(CONFIG_TCG_INTERPRETER)
+
+Int128 int128_divu(Int128 a_s, Int128 b_s)
+{
+ Int128Alias r, a, b;
+
+ a.s = a_s;
+ b.s = b_s;
+ r.u = a.u / b.u;
+ return r.s;
+}
+
+Int128 int128_remu(Int128 a_s, Int128 b_s)
+{
+ Int128Alias r, a, b;
+
+ a.s = a_s;
+ b.s = b_s;
+ r.u = a.u % b.u;
+ return r.s;
+}
+
+Int128 int128_divs(Int128 a_s, Int128 b_s)
+{
+ Int128Alias r, a, b;
+
+ a.s = a_s;
+ b.s = b_s;
+ r.i = a.i / b.i;
+ return r.s;
+}
+
+Int128 int128_rems(Int128 a_s, Int128 b_s)
+{
+ Int128Alias r, a, b;
+
+ a.s = a_s;
+ b.s = b_s;
+ r.i = a.i % b.i;
+ return r.s;
+}
+
#endif
--
2.34.1
^ permalink raw reply related [flat|nested] 43+ messages in thread* [PATCH v2 31/36] tcg/i386: Add TCG_TARGET_CALL_{RET,ARG}_I128
2022-10-21 7:15 [PATCH v2 00/36] tcg: Support for Int128 with helpers Richard Henderson
` (29 preceding siblings ...)
2022-10-21 7:15 ` [PATCH v2 30/36] include/qemu/int128: Use Int128 structure for TCI Richard Henderson
@ 2022-10-21 7:15 ` Richard Henderson
2022-10-21 7:15 ` [PATCH v2 32/36] tcg/tci: Fix big-endian return register ordering Richard Henderson
` (4 subsequent siblings)
35 siblings, 0 replies; 43+ messages in thread
From: Richard Henderson @ 2022-10-21 7:15 UTC (permalink / raw)
To: qemu-devel
Fill in the parameters for the host ABI for Int128.
Adjust tcg_target_call_oarg_reg for _WIN64, and
tcg_out_call for i386 sysv. Allow TCG_TYPE_V128
stores without AVX enabled.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
tcg/i386/tcg-target.h | 10 ++++++++++
tcg/i386/tcg-target.c.inc | 32 +++++++++++++++++++++++++++++++-
2 files changed, 41 insertions(+), 1 deletion(-)
diff --git a/tcg/i386/tcg-target.h b/tcg/i386/tcg-target.h
index 7edb7f1d9a..9e0e82d80a 100644
--- a/tcg/i386/tcg-target.h
+++ b/tcg/i386/tcg-target.h
@@ -100,6 +100,16 @@ typedef enum {
#endif
#define TCG_TARGET_CALL_ARG_I32 TCG_CALL_ARG_NORMAL
#define TCG_TARGET_CALL_ARG_I64 TCG_CALL_ARG_NORMAL
+#if defined(_WIN64)
+# define TCG_TARGET_CALL_ARG_I128 TCG_CALL_ARG_BY_REF
+# define TCG_TARGET_CALL_RET_I128 TCG_CALL_RET_BY_VEC
+#elif TCG_TARGET_REG_BITS == 64
+# define TCG_TARGET_CALL_ARG_I128 TCG_CALL_ARG_NORMAL
+# define TCG_TARGET_CALL_RET_I128 TCG_CALL_RET_NORMAL
+#else
+# define TCG_TARGET_CALL_ARG_I128 TCG_CALL_ARG_NORMAL
+# define TCG_TARGET_CALL_RET_I128 TCG_CALL_RET_BY_REF
+#endif
extern bool have_bmi1;
extern bool have_popcnt;
diff --git a/tcg/i386/tcg-target.c.inc b/tcg/i386/tcg-target.c.inc
index 82c8491152..3b0c567270 100644
--- a/tcg/i386/tcg-target.c.inc
+++ b/tcg/i386/tcg-target.c.inc
@@ -120,6 +120,13 @@ static TCGReg tcg_target_call_oarg_reg(TCGCallReturnKind kind, int slot)
return TCG_REG_EDX;
}
break;
+#ifdef _WIN64
+ case TCG_CALL_RET_BY_VEC:
+ if (slot == 0) {
+ return TCG_REG_XMM0;
+ }
+ break;
+#endif
default:
break;
}
@@ -1194,9 +1201,16 @@ static void tcg_out_st(TCGContext *s, TCGType type, TCGReg arg,
* The gvec infrastructure is asserts that v128 vector loads
* and stores use a 16-byte aligned offset. Validate that the
* final pointer is aligned by using an insn that will SIGSEGV.
+ *
+ * This specific instance is also used by TCG_CALL_RET_BY_VEC,
+ * for _WIN64, which must have SSE2 but may not have AVX.
*/
tcg_debug_assert(arg >= 16);
- tcg_out_vex_modrm_offset(s, OPC_MOVDQA_WxVx, arg, 0, arg1, arg2);
+ if (have_avx1) {
+ tcg_out_vex_modrm_offset(s, OPC_MOVDQA_WxVx, arg, 0, arg1, arg2);
+ } else {
+ tcg_out_modrm_offset(s, OPC_MOVDQA_WxVx, arg, arg1, arg2);
+ }
break;
case TCG_TYPE_V256:
/*
@@ -1683,6 +1697,22 @@ static void tcg_out_call(TCGContext *s, const tcg_insn_unit *dest,
const TCGHelperInfo *info)
{
tcg_out_branch(s, 1, dest);
+
+#ifndef _WIN32
+ if (TCG_TARGET_REG_BITS == 32 && info->out_kind == TCG_CALL_RET_BY_REF) {
+ /*
+ * The sysv i386 abi for struct return places a reference as the
+ * first argument of the stack, and pops that argument with the
+ * return statement. Since we want to retain the aligned stack
+ * pointer for the callee, we do not want to actually push that
+ * argument before the call but rely on the normal store to the
+ * stack slot. But we do need to compensate for the pop in order
+ * to reset our correct stack pointer value.
+ * Pushing a garbage value back onto the stack is quickest.
+ */
+ tcg_out_push(s, TCG_REG_EAX);
+ }
+#endif
}
static void tcg_out_jmp(TCGContext *s, const tcg_insn_unit *dest)
--
2.34.1
^ permalink raw reply related [flat|nested] 43+ messages in thread* [PATCH v2 32/36] tcg/tci: Fix big-endian return register ordering
2022-10-21 7:15 [PATCH v2 00/36] tcg: Support for Int128 with helpers Richard Henderson
` (30 preceding siblings ...)
2022-10-21 7:15 ` [PATCH v2 31/36] tcg/i386: Add TCG_TARGET_CALL_{RET,ARG}_I128 Richard Henderson
@ 2022-10-21 7:15 ` Richard Henderson
2022-10-21 7:15 ` [PATCH v2 33/36] tcg/tci: Add TCG_TARGET_CALL_{RET,ARG}_I128 Richard Henderson
` (3 subsequent siblings)
35 siblings, 0 replies; 43+ messages in thread
From: Richard Henderson @ 2022-10-21 7:15 UTC (permalink / raw)
To: qemu-devel
We expect the backend to require register pairs in
host-endian ordering, thus for big-endian the first
register of a pair contains the high part.
We were forcing R0 to contain the low part for calls.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
tcg/tci.c | 21 +++++++++++----------
1 file changed, 11 insertions(+), 10 deletions(-)
diff --git a/tcg/tci.c b/tcg/tci.c
index bdfac83492..114dece206 100644
--- a/tcg/tci.c
+++ b/tcg/tci.c
@@ -521,27 +521,28 @@ uintptr_t QEMU_DISABLE_CFI tcg_qemu_tb_exec(CPUArchState *env,
ffi_call(pptr[1], pptr[0], stack, call_slots);
}
- /* Any result winds up "left-aligned" in the stack[0] slot. */
switch (len) {
case 0: /* void */
break;
case 1: /* uint32_t */
/*
+ * The result winds up "left-aligned" in the stack[0] slot.
* Note that libffi has an odd special case in that it will
* always widen an integral result to ffi_arg.
*/
- if (sizeof(ffi_arg) == 4) {
- regs[TCG_REG_R0] = *(uint32_t *)stack;
- break;
- }
- /* fall through */
- case 2: /* uint64_t */
- if (TCG_TARGET_REG_BITS == 32) {
- tci_write_reg64(regs, TCG_REG_R1, TCG_REG_R0, stack[0]);
+ if (sizeof(ffi_arg) == 8) {
+ regs[TCG_REG_R0] = (uint32_t)stack[0];
} else {
- regs[TCG_REG_R0] = stack[0];
+ regs[TCG_REG_R0] = *(uint32_t *)stack;
}
break;
+ case 2: /* uint64_t */
+ /*
+ * For TCG_TARGET_REG_BITS == 32, the register pair
+ * must stay in host memory order.
+ */
+ memcpy(®s[TCG_REG_R0], stack, 8);
+ break;
default:
g_assert_not_reached();
}
--
2.34.1
^ permalink raw reply related [flat|nested] 43+ messages in thread* [PATCH v2 33/36] tcg/tci: Add TCG_TARGET_CALL_{RET,ARG}_I128
2022-10-21 7:15 [PATCH v2 00/36] tcg: Support for Int128 with helpers Richard Henderson
` (31 preceding siblings ...)
2022-10-21 7:15 ` [PATCH v2 32/36] tcg/tci: Fix big-endian return register ordering Richard Henderson
@ 2022-10-21 7:15 ` Richard Henderson
2022-10-21 10:47 ` Philippe Mathieu-Daudé
2022-10-21 7:15 ` [PATCH v2 34/36] tcg: " Richard Henderson
` (2 subsequent siblings)
35 siblings, 1 reply; 43+ messages in thread
From: Richard Henderson @ 2022-10-21 7:15 UTC (permalink / raw)
To: qemu-devel
Fill in the parameters for libffi for Int128.
Adjust the interpreter to allow for 16-byte return values.
Adjust tcg_out_call to record the return value length.
Call parameters are no longer all the same size, so we
cannot reuse the same call_slots array for every function.
Compute it each time now, but only fill in slots required
for the call we're about to make.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
tcg/tci/tcg-target.h | 3 +++
tcg/tcg.c | 18 ++++++++++++++++
tcg/tci.c | 45 ++++++++++++++++++++--------------------
tcg/tci/tcg-target.c.inc | 8 +++----
4 files changed, 48 insertions(+), 26 deletions(-)
diff --git a/tcg/tci/tcg-target.h b/tcg/tci/tcg-target.h
index 94ec541b4e..9d569c9e04 100644
--- a/tcg/tci/tcg-target.h
+++ b/tcg/tci/tcg-target.h
@@ -161,10 +161,13 @@ typedef enum {
#if TCG_TARGET_REG_BITS == 32
# define TCG_TARGET_CALL_ARG_I32 TCG_CALL_ARG_EVEN
# define TCG_TARGET_CALL_ARG_I64 TCG_CALL_ARG_EVEN
+# define TCG_TARGET_CALL_ARG_I128 TCG_CALL_ARG_EVEN
#else
# define TCG_TARGET_CALL_ARG_I32 TCG_CALL_ARG_NORMAL
# define TCG_TARGET_CALL_ARG_I64 TCG_CALL_ARG_NORMAL
+# define TCG_TARGET_CALL_ARG_I128 TCG_CALL_ARG_NORMAL
#endif
+#define TCG_TARGET_CALL_RET_I128 TCG_CALL_RET_NORMAL
#define HAVE_TCG_QEMU_TB_EXEC
#define TCG_TARGET_NEED_POOL_LABELS
diff --git a/tcg/tcg.c b/tcg/tcg.c
index 9ec5b85f44..f921a5ca24 100644
--- a/tcg/tcg.c
+++ b/tcg/tcg.c
@@ -548,6 +548,22 @@ static GHashTable *helper_table;
#ifdef CONFIG_TCG_INTERPRETER
static ffi_type *typecode_to_ffi(int argmask)
{
+ /*
+ * libffi does not support __int128_t, so we have forced Int128
+ * to use the structure definition instead of the builtin type.
+ */
+ static ffi_type *ffi_type_i128_elements[3] = {
+ &ffi_type_uint64,
+ &ffi_type_uint64,
+ NULL
+ };
+ static ffi_type ffi_type_i128 = {
+ .size = 16,
+ .alignment = __alignof__(Int128),
+ .type = FFI_TYPE_STRUCT,
+ .elements = ffi_type_i128_elements,
+ };
+
switch (argmask) {
case dh_typecode_void:
return &ffi_type_void;
@@ -561,6 +577,8 @@ static ffi_type *typecode_to_ffi(int argmask)
return &ffi_type_sint64;
case dh_typecode_ptr:
return &ffi_type_pointer;
+ case dh_typecode_i128:
+ return &ffi_type_i128;
}
g_assert_not_reached();
}
diff --git a/tcg/tci.c b/tcg/tci.c
index 114dece206..9154a3f22b 100644
--- a/tcg/tci.c
+++ b/tcg/tci.c
@@ -471,12 +471,9 @@ uintptr_t QEMU_DISABLE_CFI tcg_qemu_tb_exec(CPUArchState *env,
tcg_target_ulong regs[TCG_TARGET_NB_REGS];
uint64_t stack[(TCG_STATIC_CALL_ARGS_SIZE + TCG_STATIC_FRAME_SIZE)
/ sizeof(uint64_t)];
- void *call_slots[TCG_STATIC_CALL_ARGS_SIZE / sizeof(uint64_t)];
regs[TCG_AREG0] = (tcg_target_ulong)env;
regs[TCG_REG_CALL_STACK] = (uintptr_t)stack;
- /* Other call_slots entries initialized at first use (see below). */
- call_slots[0] = NULL;
tci_assert(tb_ptr);
for (;;) {
@@ -499,26 +496,27 @@ uintptr_t QEMU_DISABLE_CFI tcg_qemu_tb_exec(CPUArchState *env,
switch (opc) {
case INDEX_op_call:
- /*
- * Set up the ffi_avalue array once, delayed until now
- * because many TB's do not make any calls. In tcg_gen_callN,
- * we arranged for every real argument to be "left-aligned"
- * in each 64-bit slot.
- */
- if (unlikely(call_slots[0] == NULL)) {
- for (int i = 0; i < ARRAY_SIZE(call_slots); ++i) {
- call_slots[i] = &stack[i];
- }
- }
-
- tci_args_nl(insn, tb_ptr, &len, &ptr);
-
- /* Helper functions may need to access the "return address" */
- tci_tb_ptr = (uintptr_t)tb_ptr;
-
{
- void **pptr = ptr;
- ffi_call(pptr[1], pptr[0], stack, call_slots);
+ void *call_slots[MAX_OPC_PARAM_IARGS];
+ ffi_cif *cif;
+ void *func;
+ unsigned i, s, n;
+
+ tci_args_nl(insn, tb_ptr, &len, &ptr);
+ func = ((void **)ptr)[0];
+ cif = ((void **)ptr)[1];
+
+ n = cif->nargs;
+ tci_assert(n <= MAX_OPC_PARAM_IARGS);
+ for (i = s = 0; i < n; ++i) {
+ ffi_type *t = cif->arg_types[i];
+ call_slots[i] = &stack[s];
+ s += DIV_ROUND_UP(t->size, 8);
+ }
+
+ /* Helper functions may need to access the "return address" */
+ tci_tb_ptr = (uintptr_t)tb_ptr;
+ ffi_call(cif, func, stack, call_slots);
}
switch (len) {
@@ -543,6 +541,9 @@ uintptr_t QEMU_DISABLE_CFI tcg_qemu_tb_exec(CPUArchState *env,
*/
memcpy(®s[TCG_REG_R0], stack, 8);
break;
+ case 3: /* Int128 */
+ memcpy(®s[TCG_REG_R0], stack, 16);
+ break;
default:
g_assert_not_reached();
}
diff --git a/tcg/tci/tcg-target.c.inc b/tcg/tci/tcg-target.c.inc
index 914806c216..a20a983422 100644
--- a/tcg/tci/tcg-target.c.inc
+++ b/tcg/tci/tcg-target.c.inc
@@ -576,11 +576,11 @@ static void tcg_out_call(TCGContext *s, const tcg_insn_unit *func,
if (cif->rtype == &ffi_type_void) {
which = 0;
- } else if (cif->rtype->size == 4) {
- which = 1;
} else {
- tcg_debug_assert(cif->rtype->size == 8);
- which = 2;
+ tcg_debug_assert(cif->rtype->size == 4 ||
+ cif->rtype->size == 8 ||
+ cif->rtype->size == 16);
+ which = ctz32(cif->rtype->size) - 1;
}
new_pool_l2(s, 20, s->code_ptr, 0, (uintptr_t)func, (uintptr_t)cif);
insn = deposit32(insn, 0, 8, INDEX_op_call);
--
2.34.1
^ permalink raw reply related [flat|nested] 43+ messages in thread* Re: [PATCH v2 33/36] tcg/tci: Add TCG_TARGET_CALL_{RET,ARG}_I128
2022-10-21 7:15 ` [PATCH v2 33/36] tcg/tci: Add TCG_TARGET_CALL_{RET,ARG}_I128 Richard Henderson
@ 2022-10-21 10:47 ` Philippe Mathieu-Daudé
2022-10-22 3:48 ` Richard Henderson
0 siblings, 1 reply; 43+ messages in thread
From: Philippe Mathieu-Daudé @ 2022-10-21 10:47 UTC (permalink / raw)
To: Richard Henderson, qemu-devel
On 21/10/22 09:15, Richard Henderson wrote:
> Fill in the parameters for libffi for Int128.
> Adjust the interpreter to allow for 16-byte return values.
> Adjust tcg_out_call to record the return value length.
>
> Call parameters are no longer all the same size, so we
> cannot reuse the same call_slots array for every function.
> Compute it each time now, but only fill in slots required
> for the call we're about to make.
>
> Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
> ---
> tcg/tci/tcg-target.h | 3 +++
> tcg/tcg.c | 18 ++++++++++++++++
> tcg/tci.c | 45 ++++++++++++++++++++--------------------
> tcg/tci/tcg-target.c.inc | 8 +++----
> 4 files changed, 48 insertions(+), 26 deletions(-)
> static ffi_type *typecode_to_ffi(int argmask)
> {
> + /*
> + * libffi does not support __int128_t, so we have forced Int128
> + * to use the structure definition instead of the builtin type.
> + */
> + static ffi_type *ffi_type_i128_elements[3] = {
static const.
> + &ffi_type_uint64,
> + &ffi_type_uint64,
> + NULL
> + };
> + static ffi_type ffi_type_i128 = {
static const.
> + .size = 16,
> + .alignment = __alignof__(Int128),
> + .type = FFI_TYPE_STRUCT,
> + .elements = ffi_type_i128_elements,
> + };
> @@ -499,26 +496,27 @@ uintptr_t QEMU_DISABLE_CFI tcg_qemu_tb_exec(CPUArchState *env,
>
> switch (opc) {
> case INDEX_op_call:
> - /*
> - * Set up the ffi_avalue array once, delayed until now
> - * because many TB's do not make any calls. In tcg_gen_callN,
> - * we arranged for every real argument to be "left-aligned"
> - * in each 64-bit slot.
> - */
> - if (unlikely(call_slots[0] == NULL)) {
> - for (int i = 0; i < ARRAY_SIZE(call_slots); ++i) {
> - call_slots[i] = &stack[i];
> - }
> - }
> -
> - tci_args_nl(insn, tb_ptr, &len, &ptr);
> -
> - /* Helper functions may need to access the "return address" */
> - tci_tb_ptr = (uintptr_t)tb_ptr;
> -
> {
> - void **pptr = ptr;
> - ffi_call(pptr[1], pptr[0], stack, call_slots);
> + void *call_slots[MAX_OPC_PARAM_IARGS];
> + ffi_cif *cif;
> + void *func;
> + unsigned i, s, n;
> +
> + tci_args_nl(insn, tb_ptr, &len, &ptr);
> + func = ((void **)ptr)[0];
> + cif = ((void **)ptr)[1];
> +
> + n = cif->nargs;
> + tci_assert(n <= MAX_OPC_PARAM_IARGS);
> + for (i = s = 0; i < n; ++i) {
> + ffi_type *t = cif->arg_types[i];
> + call_slots[i] = &stack[s];
> + s += DIV_ROUND_UP(t->size, 8);
> + }
> +
> + /* Helper functions may need to access the "return address" */
> + tci_tb_ptr = (uintptr_t)tb_ptr;
> + ffi_call(cif, func, stack, call_slots);
> }
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
^ permalink raw reply [flat|nested] 43+ messages in thread* Re: [PATCH v2 33/36] tcg/tci: Add TCG_TARGET_CALL_{RET,ARG}_I128
2022-10-21 10:47 ` Philippe Mathieu-Daudé
@ 2022-10-22 3:48 ` Richard Henderson
0 siblings, 0 replies; 43+ messages in thread
From: Richard Henderson @ 2022-10-22 3:48 UTC (permalink / raw)
To: Philippe Mathieu-Daudé, qemu-devel
On 10/21/22 20:47, Philippe Mathieu-Daudé wrote:
> On 21/10/22 09:15, Richard Henderson wrote:
>> Fill in the parameters for libffi for Int128.
>> Adjust the interpreter to allow for 16-byte return values.
>> Adjust tcg_out_call to record the return value length.
>>
>> Call parameters are no longer all the same size, so we
>> cannot reuse the same call_slots array for every function.
>> Compute it each time now, but only fill in slots required
>> for the call we're about to make.
>>
>> Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
>> ---
>> tcg/tci/tcg-target.h | 3 +++
>> tcg/tcg.c | 18 ++++++++++++++++
>> tcg/tci.c | 45 ++++++++++++++++++++--------------------
>> tcg/tci/tcg-target.c.inc | 8 +++----
>> 4 files changed, 48 insertions(+), 26 deletions(-)
>
>> static ffi_type *typecode_to_ffi(int argmask)
>> {
>> + /*
>> + * libffi does not support __int128_t, so we have forced Int128
>> + * to use the structure definition instead of the builtin type.
>> + */
>> + static ffi_type *ffi_type_i128_elements[3] = {
>
> static const.
>
>> + &ffi_type_uint64,
>> + &ffi_type_uint64,
>> + NULL
>> + };
>> + static ffi_type ffi_type_i128 = {
>
> static const.
Can't, because of the libffi interface. Indeed this second structure will be modified by
the library to complete the library-specific initialization.
r~
^ permalink raw reply [flat|nested] 43+ messages in thread
* [PATCH v2 34/36] tcg: Add TCG_TARGET_CALL_{RET,ARG}_I128
2022-10-21 7:15 [PATCH v2 00/36] tcg: Support for Int128 with helpers Richard Henderson
` (32 preceding siblings ...)
2022-10-21 7:15 ` [PATCH v2 33/36] tcg/tci: Add TCG_TARGET_CALL_{RET,ARG}_I128 Richard Henderson
@ 2022-10-21 7:15 ` Richard Henderson
2022-10-21 7:15 ` [PATCH v2 35/36] tcg: Add temp allocation for TCGv_i128 Richard Henderson
2022-10-21 7:15 ` [PATCH v2 36/36] tcg: Add tcg_gen_extr_i128_i64, tcg_gen_concat_i64_i128 Richard Henderson
35 siblings, 0 replies; 43+ messages in thread
From: Richard Henderson @ 2022-10-21 7:15 UTC (permalink / raw)
To: qemu-devel
Fill in the parameters for the host ABI for Int128 for
those backends which require no extra modification.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
tcg/aarch64/tcg-target.h | 2 ++
tcg/arm/tcg-target.h | 2 ++
tcg/loongarch64/tcg-target.h | 2 ++
tcg/mips/tcg-target.h | 2 ++
tcg/riscv/tcg-target.h | 3 +++
tcg/s390x/tcg-target.h | 2 ++
tcg/sparc64/tcg-target.h | 2 ++
tcg/tcg.c | 11 ++++++-----
tcg/ppc/tcg-target.c.inc | 3 +++
9 files changed, 24 insertions(+), 5 deletions(-)
diff --git a/tcg/aarch64/tcg-target.h b/tcg/aarch64/tcg-target.h
index 413a5410c5..0dff5807f6 100644
--- a/tcg/aarch64/tcg-target.h
+++ b/tcg/aarch64/tcg-target.h
@@ -54,6 +54,8 @@ typedef enum {
#define TCG_TARGET_CALL_STACK_OFFSET 0
#define TCG_TARGET_CALL_ARG_I32 TCG_CALL_ARG_NORMAL
#define TCG_TARGET_CALL_ARG_I64 TCG_CALL_ARG_NORMAL
+#define TCG_TARGET_CALL_ARG_I128 TCG_CALL_ARG_NORMAL
+#define TCG_TARGET_CALL_RET_I128 TCG_CALL_RET_NORMAL
/* optional instructions */
#define TCG_TARGET_HAS_div_i32 1
diff --git a/tcg/arm/tcg-target.h b/tcg/arm/tcg-target.h
index b7843d2d54..6613d3d791 100644
--- a/tcg/arm/tcg-target.h
+++ b/tcg/arm/tcg-target.h
@@ -91,6 +91,8 @@ extern bool use_neon_instructions;
#define TCG_TARGET_CALL_STACK_OFFSET 0
#define TCG_TARGET_CALL_ARG_I32 TCG_CALL_ARG_NORMAL
#define TCG_TARGET_CALL_ARG_I64 TCG_CALL_ARG_EVEN
+#define TCG_TARGET_CALL_ARG_I128 TCG_CALL_ARG_EVEN
+#define TCG_TARGET_CALL_RET_I128 TCG_CALL_RET_BY_REF
/* optional instructions */
#define TCG_TARGET_HAS_ext8s_i32 1
diff --git a/tcg/loongarch64/tcg-target.h b/tcg/loongarch64/tcg-target.h
index 74fbb22dfd..54e3478355 100644
--- a/tcg/loongarch64/tcg-target.h
+++ b/tcg/loongarch64/tcg-target.h
@@ -91,6 +91,8 @@ typedef enum {
#define TCG_TARGET_CALL_STACK_OFFSET 0
#define TCG_TARGET_CALL_ARG_I32 TCG_CALL_ARG_NORMAL
#define TCG_TARGET_CALL_ARG_I64 TCG_CALL_ARG_NORMAL
+#define TCG_TARGET_CALL_ARG_I128 TCG_CALL_ARG_NORMAL
+#define TCG_TARGET_CALL_RET_I128 TCG_CALL_RET_NORMAL
/* optional instructions */
#define TCG_TARGET_HAS_movcond_i32 0
diff --git a/tcg/mips/tcg-target.h b/tcg/mips/tcg-target.h
index 15721c3e42..b235cba8ba 100644
--- a/tcg/mips/tcg-target.h
+++ b/tcg/mips/tcg-target.h
@@ -89,6 +89,8 @@ typedef enum {
# define TCG_TARGET_CALL_ARG_I64 TCG_CALL_ARG_NORMAL
#endif
#define TCG_TARGET_CALL_ARG_I32 TCG_CALL_ARG_NORMAL
+#define TCG_TARGET_CALL_ARG_I128 TCG_CALL_ARG_EVEN
+#define TCG_TARGET_CALL_RET_I128 TCG_CALL_RET_NORMAL
/* MOVN/MOVZ instructions detection */
#if (defined(__mips_isa_rev) && (__mips_isa_rev >= 1)) || \
diff --git a/tcg/riscv/tcg-target.h b/tcg/riscv/tcg-target.h
index 232537ccea..d61ca902d3 100644
--- a/tcg/riscv/tcg-target.h
+++ b/tcg/riscv/tcg-target.h
@@ -85,9 +85,12 @@ typedef enum {
#define TCG_TARGET_CALL_ARG_I32 TCG_CALL_ARG_NORMAL
#if TCG_TARGET_REG_BITS == 32
#define TCG_TARGET_CALL_ARG_I64 TCG_CALL_ARG_EVEN
+#define TCG_TARGET_CALL_ARG_I128 TCG_CALL_ARG_EVEN
#else
#define TCG_TARGET_CALL_ARG_I64 TCG_CALL_ARG_NORMAL
+#define TCG_TARGET_CALL_ARG_I128 TCG_CALL_ARG_NORMAL
#endif
+#define TCG_TARGET_CALL_RET_I128 TCG_CALL_RET_NORMAL
/* optional instructions */
#define TCG_TARGET_HAS_movcond_i32 0
diff --git a/tcg/s390x/tcg-target.h b/tcg/s390x/tcg-target.h
index db5665c375..9a3856f0b3 100644
--- a/tcg/s390x/tcg-target.h
+++ b/tcg/s390x/tcg-target.h
@@ -168,6 +168,8 @@ extern uint64_t s390_facilities[3];
#define TCG_TARGET_CALL_STACK_OFFSET 160
#define TCG_TARGET_CALL_ARG_I32 TCG_CALL_ARG_EXTEND
#define TCG_TARGET_CALL_ARG_I64 TCG_CALL_RET_NORMAL
+#define TCG_TARGET_CALL_ARG_I128 TCG_CALL_ARG_BY_REF
+#define TCG_TARGET_CALL_RET_I128 TCG_CALL_RET_BY_REF
#define TCG_TARGET_HAS_MEMORY_BSWAP 1
diff --git a/tcg/sparc64/tcg-target.h b/tcg/sparc64/tcg-target.h
index 0044ac8d78..53cfa843da 100644
--- a/tcg/sparc64/tcg-target.h
+++ b/tcg/sparc64/tcg-target.h
@@ -73,6 +73,8 @@ typedef enum {
#define TCG_TARGET_CALL_STACK_OFFSET (128 + 6*8 + TCG_TARGET_STACK_BIAS)
#define TCG_TARGET_CALL_ARG_I32 TCG_CALL_ARG_EXTEND
#define TCG_TARGET_CALL_ARG_I64 TCG_CALL_ARG_NORMAL
+#define TCG_TARGET_CALL_ARG_I128 TCG_CALL_ARG_NORMAL
+#define TCG_TARGET_CALL_RET_I128 TCG_CALL_RET_NORMAL
#if defined(__VIS__) && __VIS__ >= 0x300
#define use_vis3_instructions 1
diff --git a/tcg/tcg.c b/tcg/tcg.c
index f921a5ca24..7cde82f12d 100644
--- a/tcg/tcg.c
+++ b/tcg/tcg.c
@@ -760,12 +760,13 @@ static void init_call_layout(TCGHelperInfo *info)
break;
case dh_typecode_i128:
/*
- * No matter the call return method, we must have all of
- * the temp subindexes in the call for liveness.
+ * For 32-bit hosts, do not split out individual temps.
+ * For 64-bit hosts, we need to add both temps for liveness
+ * no matter the backend calling convention.
*/
info->nr_out = TCG_TARGET_REG_BITS == 32 ? 1 : 2;
- info->out_kind = TCG_CALL_RET_NORMAL; /* TODO */
- switch (/* TODO */ TCG_CALL_RET_NORMAL) {
+ info->out_kind = TCG_TARGET_CALL_RET_I128;
+ switch (TCG_TARGET_CALL_RET_I128) {
case TCG_CALL_RET_NORMAL:
if (TCG_TARGET_REG_BITS == 32) {
info->out_kind = TCG_CALL_RET_NORMAL_4;
@@ -827,7 +828,7 @@ static void init_call_layout(TCGHelperInfo *info)
break;
case dh_typecode_i128:
type = TCG_TYPE_I128;
- kind = TCG_CALL_ARG_NORMAL; /* TODO */
+ kind = TCG_TARGET_CALL_ARG_I128;
break;
default:
g_assert_not_reached();
diff --git a/tcg/ppc/tcg-target.c.inc b/tcg/ppc/tcg-target.c.inc
index 781ecfe161..e86d4a5e78 100644
--- a/tcg/ppc/tcg-target.c.inc
+++ b/tcg/ppc/tcg-target.c.inc
@@ -54,6 +54,9 @@
#else
# define TCG_TARGET_CALL_ARG_I64 TCG_CALL_ARG_NORMAL
#endif
+/* Note sysv arg alignment applies only to 2-word types, not more. */
+#define TCG_TARGET_CALL_ARG_I128 TCG_CALL_ARG_NORMAL
+#define TCG_TARGET_CALL_RET_I128 TCG_CALL_RET_NORMAL
/* For some memory operations, we need a scratch that isn't R0. For the AIX
calling convention, we can re-use the TOC register since we'll be reloading
--
2.34.1
^ permalink raw reply related [flat|nested] 43+ messages in thread* [PATCH v2 35/36] tcg: Add temp allocation for TCGv_i128
2022-10-21 7:15 [PATCH v2 00/36] tcg: Support for Int128 with helpers Richard Henderson
` (33 preceding siblings ...)
2022-10-21 7:15 ` [PATCH v2 34/36] tcg: " Richard Henderson
@ 2022-10-21 7:15 ` Richard Henderson
2022-10-21 7:15 ` [PATCH v2 36/36] tcg: Add tcg_gen_extr_i128_i64, tcg_gen_concat_i64_i128 Richard Henderson
35 siblings, 0 replies; 43+ messages in thread
From: Richard Henderson @ 2022-10-21 7:15 UTC (permalink / raw)
To: qemu-devel
This enables allocation of i128. The type is not yet
usable, as we have not yet added data movement ops.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
include/tcg/tcg.h | 32 ++++++++++++++++++++++++++++++++
tcg/tcg.c | 32 +++++++++++++++++++++++++++-----
2 files changed, 59 insertions(+), 5 deletions(-)
diff --git a/include/tcg/tcg.h b/include/tcg/tcg.h
index 88fb4f1d17..db2e10610d 100644
--- a/include/tcg/tcg.h
+++ b/include/tcg/tcg.h
@@ -701,6 +701,11 @@ static inline TCGTemp *tcgv_i64_temp(TCGv_i64 v)
return tcgv_i32_temp((TCGv_i32)v);
}
+static inline TCGTemp *tcgv_i128_temp(TCGv_i128 v)
+{
+ return tcgv_i32_temp((TCGv_i32)v);
+}
+
static inline TCGTemp *tcgv_ptr_temp(TCGv_ptr v)
{
return tcgv_i32_temp((TCGv_i32)v);
@@ -721,6 +726,11 @@ static inline TCGArg tcgv_i64_arg(TCGv_i64 v)
return temp_arg(tcgv_i64_temp(v));
}
+static inline TCGArg tcgv_i128_arg(TCGv_i128 v)
+{
+ return temp_arg(tcgv_i128_temp(v));
+}
+
static inline TCGArg tcgv_ptr_arg(TCGv_ptr v)
{
return temp_arg(tcgv_ptr_temp(v));
@@ -742,6 +752,11 @@ static inline TCGv_i64 temp_tcgv_i64(TCGTemp *t)
return (TCGv_i64)temp_tcgv_i32(t);
}
+static inline TCGv_i128 temp_tcgv_i128(TCGTemp *t)
+{
+ return (TCGv_i128)temp_tcgv_i32(t);
+}
+
static inline TCGv_ptr temp_tcgv_ptr(TCGTemp *t)
{
return (TCGv_ptr)temp_tcgv_i32(t);
@@ -864,6 +879,11 @@ static inline void tcg_temp_free_i64(TCGv_i64 arg)
tcg_temp_free_internal(tcgv_i64_temp(arg));
}
+static inline void tcg_temp_free_i128(TCGv_i128 arg)
+{
+ tcg_temp_free_internal(tcgv_i128_temp(arg));
+}
+
static inline void tcg_temp_free_ptr(TCGv_ptr arg)
{
tcg_temp_free_internal(tcgv_ptr_temp(arg));
@@ -912,6 +932,18 @@ static inline TCGv_i64 tcg_temp_local_new_i64(void)
return temp_tcgv_i64(t);
}
+static inline TCGv_i128 tcg_temp_new_i128(void)
+{
+ TCGTemp *t = tcg_temp_new_internal(TCG_TYPE_I128, false);
+ return temp_tcgv_i128(t);
+}
+
+static inline TCGv_i128 tcg_temp_local_new_i128(void)
+{
+ TCGTemp *t = tcg_temp_new_internal(TCG_TYPE_I128, true);
+ return temp_tcgv_i128(t);
+}
+
static inline TCGv_ptr tcg_global_mem_new_ptr(TCGv_ptr reg, intptr_t offset,
const char *name)
{
diff --git a/tcg/tcg.c b/tcg/tcg.c
index 7cde82f12d..adf592ac96 100644
--- a/tcg/tcg.c
+++ b/tcg/tcg.c
@@ -1250,18 +1250,39 @@ TCGTemp *tcg_temp_new_internal(TCGType type, bool temp_local)
tcg_debug_assert(ts->base_type == type);
tcg_debug_assert(ts->kind == kind);
} else {
+ bool want_pair = false;
+ TCGType half = type;
+
+ switch (type) {
+ case TCG_TYPE_I32:
+ case TCG_TYPE_V64:
+ case TCG_TYPE_V128:
+ case TCG_TYPE_V256:
+ break;
+ case TCG_TYPE_I64:
+ half = TCG_TYPE_I32;
+ want_pair = TCG_TARGET_REG_BITS == 32;
+ break;
+ case TCG_TYPE_I128:
+ half = TCG_TYPE_I64;
+ want_pair = TCG_TARGET_REG_BITS == 64;
+ break;
+ default:
+ g_assert_not_reached();
+ }
+
ts = tcg_temp_alloc(s);
- if (TCG_TARGET_REG_BITS == 32 && type == TCG_TYPE_I64) {
+ if (want_pair) {
TCGTemp *ts2 = tcg_temp_alloc(s);
ts->base_type = type;
- ts->type = TCG_TYPE_I32;
+ ts->type = half;
ts->temp_allocated = 1;
ts->kind = kind;
tcg_debug_assert(ts2 == ts + 1);
- ts2->base_type = TCG_TYPE_I64;
- ts2->type = TCG_TYPE_I32;
+ ts2->base_type = type;
+ ts2->type = half;
ts2->temp_allocated = 1;
ts2->temp_subindex = 1;
ts2->kind = kind;
@@ -2773,7 +2794,7 @@ static void la_cross_call(TCGContext *s, int nt)
for (i = 0; i < nt; i++) {
TCGTemp *ts = &s->temps[i];
- if (!(ts->state & TS_DEAD)) {
+ if (ts->type != TCG_TYPE_I128 && !(ts->state & TS_DEAD)) {
TCGRegSet *pset = la_temp_pref(ts);
TCGRegSet set = *pset;
@@ -3404,6 +3425,7 @@ static void temp_allocate_frame(TCGContext *s, TCGTemp *ts)
case TCG_TYPE_V64:
align = 8;
break;
+ case TCG_TYPE_I128:
case TCG_TYPE_V128:
case TCG_TYPE_V256:
/* Note that we do not require aligned storage for V256. */
--
2.34.1
^ permalink raw reply related [flat|nested] 43+ messages in thread* [PATCH v2 36/36] tcg: Add tcg_gen_extr_i128_i64, tcg_gen_concat_i64_i128
2022-10-21 7:15 [PATCH v2 00/36] tcg: Support for Int128 with helpers Richard Henderson
` (34 preceding siblings ...)
2022-10-21 7:15 ` [PATCH v2 35/36] tcg: Add temp allocation for TCGv_i128 Richard Henderson
@ 2022-10-21 7:15 ` Richard Henderson
35 siblings, 0 replies; 43+ messages in thread
From: Richard Henderson @ 2022-10-21 7:15 UTC (permalink / raw)
To: qemu-devel
Add code generation functions for data movement between
TCGv_i128 and TCGv_i64.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
include/tcg/tcg-op.h | 3 ++
include/tcg/tcg-opc.h | 4 ++
tcg/tcg-internal.h | 12 ++++++
tcg/tcg-op.c | 35 ++++++++++++++++++
tcg/tcg.c | 85 +++++++++++++++++++++++++++++++++++++++++++
5 files changed, 139 insertions(+)
diff --git a/include/tcg/tcg-op.h b/include/tcg/tcg-op.h
index 209e168305..2ce2ff28c9 100644
--- a/include/tcg/tcg-op.h
+++ b/include/tcg/tcg-op.h
@@ -735,6 +735,9 @@ void tcg_gen_extrh_i64_i32(TCGv_i32 ret, TCGv_i64 arg);
void tcg_gen_extr_i64_i32(TCGv_i32 lo, TCGv_i32 hi, TCGv_i64 arg);
void tcg_gen_extr32_i64(TCGv_i64 lo, TCGv_i64 hi, TCGv_i64 arg);
+void tcg_gen_extr_i128_i64(TCGv_i64 lo, TCGv_i64 hi, TCGv_i128 arg);
+void tcg_gen_concat_i64_i128(TCGv_i128 ret, TCGv_i64 lo, TCGv_i64 hi);
+
static inline void tcg_gen_concat32_i64(TCGv_i64 ret, TCGv_i64 lo, TCGv_i64 hi)
{
tcg_gen_deposit_i64(ret, lo, hi, 32, 32);
diff --git a/include/tcg/tcg-opc.h b/include/tcg/tcg-opc.h
index dd444734d9..6211fb3242 100644
--- a/include/tcg/tcg-opc.h
+++ b/include/tcg/tcg-opc.h
@@ -158,6 +158,10 @@ DEF(extrh_i64_i32, 1, 1, 0,
IMPL(TCG_TARGET_HAS_extrh_i64_i32)
| (TCG_TARGET_REG_BITS == 32 ? TCG_OPF_NOT_PRESENT : 0))
+/* For 32-bit host only, implemented generically using ld/st/mov. */
+DEF(extr_i128_i32, 1, 1, 1, TCG_OPF_NOT_PRESENT)
+DEF(concat4_i32_i128, 1, 4, 0, TCG_OPF_NOT_PRESENT)
+
DEF(brcond_i64, 0, 2, 2, TCG_OPF_BB_END | TCG_OPF_COND_BRANCH | IMPL64)
DEF(ext8s_i64, 1, 1, 0, IMPL64 | IMPL(TCG_TARGET_HAS_ext8s_i64))
DEF(ext16s_i64, 1, 1, 0, IMPL64 | IMPL(TCG_TARGET_HAS_ext16s_i64))
diff --git a/tcg/tcg-internal.h b/tcg/tcg-internal.h
index 44ef51ca30..8112a0a491 100644
--- a/tcg/tcg-internal.h
+++ b/tcg/tcg-internal.h
@@ -110,9 +110,21 @@ static inline TCGv_i32 TCGV_HIGH(TCGv_i64 t)
{
return temp_tcgv_i32(tcgv_i64_temp(t) + !HOST_BIG_ENDIAN);
}
+extern TCGv_i64 TCGV128_LOW(TCGv_i128)
+ QEMU_ERROR("64-bit code path is reachable");
+extern TCGv_i64 TCGV128_HIGH(TCGv_i128)
+ QEMU_ERROR("64-bit code path is reachable");
#else
extern TCGv_i32 TCGV_LOW(TCGv_i64) QEMU_ERROR("32-bit code path is reachable");
extern TCGv_i32 TCGV_HIGH(TCGv_i64) QEMU_ERROR("32-bit code path is reachable");
+static inline TCGv_i64 TCGV128_LOW(TCGv_i128 t)
+{
+ return temp_tcgv_i64(tcgv_i128_temp(t) + HOST_BIG_ENDIAN);
+}
+static inline TCGv_i64 TCGV128_HIGH(TCGv_i128 t)
+{
+ return temp_tcgv_i64(tcgv_i128_temp(t) + !HOST_BIG_ENDIAN);
+}
#endif
#endif /* TCG_INTERNAL_H */
diff --git a/tcg/tcg-op.c b/tcg/tcg-op.c
index 3ed98ffa01..6031fa89c2 100644
--- a/tcg/tcg-op.c
+++ b/tcg/tcg-op.c
@@ -2718,6 +2718,41 @@ void tcg_gen_extr32_i64(TCGv_i64 lo, TCGv_i64 hi, TCGv_i64 arg)
tcg_gen_shri_i64(hi, arg, 32);
}
+void tcg_gen_extr_i128_i64(TCGv_i64 lo, TCGv_i64 hi, TCGv_i128 arg)
+{
+ if (TCG_TARGET_REG_BITS == 32) {
+ TCGArg a_arg = tcgv_i128_arg(arg);
+ int be = HOST_BIG_ENDIAN ? 0xc : 0;
+
+ tcg_gen_op3(INDEX_op_extr_i128_i32, tcgv_i32_arg(TCGV_LOW(lo)),
+ a_arg, 0x0 ^ be);
+ tcg_gen_op3(INDEX_op_extr_i128_i32, tcgv_i32_arg(TCGV_HIGH(lo)),
+ a_arg, 0x4 ^ be);
+ tcg_gen_op3(INDEX_op_extr_i128_i32, tcgv_i32_arg(TCGV_LOW(hi)),
+ a_arg, 0x8 ^ be);
+ tcg_gen_op3(INDEX_op_extr_i128_i32, tcgv_i32_arg(TCGV_HIGH(hi)),
+ a_arg, 0xc ^ be);
+ } else {
+ tcg_gen_mov_i64(lo, TCGV128_LOW(arg));
+ tcg_gen_mov_i64(hi, TCGV128_HIGH(arg));
+ }
+}
+
+void tcg_gen_concat_i64_i128(TCGv_i128 ret, TCGv_i64 lo, TCGv_i64 hi)
+{
+ if (TCG_TARGET_REG_BITS == 32) {
+ tcg_gen_op5(INDEX_op_concat4_i32_i128,
+ tcgv_i128_arg(ret),
+ tcgv_i32_arg(TCGV_LOW(lo)),
+ tcgv_i32_arg(TCGV_HIGH(lo)),
+ tcgv_i32_arg(TCGV_LOW(hi)),
+ tcgv_i32_arg(TCGV_HIGH(hi)));
+ } else {
+ tcg_gen_mov_i64(TCGV128_LOW(ret), lo);
+ tcg_gen_mov_i64(TCGV128_HIGH(ret), hi);
+ }
+}
+
/* QEMU specific operations. */
void tcg_gen_exit_tb(const TranslationBlock *tb, unsigned idx)
diff --git a/tcg/tcg.c b/tcg/tcg.c
index adf592ac96..6974564f49 100644
--- a/tcg/tcg.c
+++ b/tcg/tcg.c
@@ -3949,6 +3949,85 @@ static void tcg_reg_alloc_mov(TCGContext *s, const TCGOp *op)
}
}
+/*
+ * Specialized code generation for TCG_TYPE_I128 on 32-bit host.
+ * Here, 128-bit values are *always* in memory, never regs or constants.
+ * Move 32-bit pieces into and out of the 128-bit memory slot.
+ */
+static void tcg_reg_alloc_exrl_i128_i32(TCGContext *s, const TCGOp *op)
+{
+ const TCGLifeData arg_life = op->life;
+ TCGTemp *ots = arg_temp(op->args[0]);
+ TCGTemp *its = arg_temp(op->args[1]);
+ TCGArg ofs = op->args[2];
+ TCGReg reg;
+
+ assert(TCG_TARGET_REG_BITS == 32);
+ tcg_debug_assert(ots->type == TCG_TYPE_I32);
+ tcg_debug_assert(!temp_readonly(ots));
+ tcg_debug_assert(its->type == TCG_TYPE_I128);
+ tcg_debug_assert(its->val_type == TEMP_VAL_MEM);
+ tcg_debug_assert(ofs < 16);
+ tcg_debug_assert((ofs & 3) == 0);
+
+ if (ots->val_type == TEMP_VAL_REG) {
+ reg = ots->reg;
+ } else {
+ reg = tcg_reg_alloc(s, tcg_target_available_regs[TCG_TYPE_I32],
+ s->reserved_regs, op->output_pref[0],
+ ots->indirect_base);
+ ots->val_type = TEMP_VAL_REG;
+ ots->reg = reg;
+ s->reg_to_temp[reg] = ots;
+ }
+
+ tcg_out_ld(s, TCG_TYPE_I32, reg,
+ its->mem_base->reg, its->mem_offset + ofs);
+ ots->mem_coherent = 0;
+
+ if (IS_DEAD_ARG(1)) {
+ temp_dead(s, its);
+ }
+}
+
+static void tcg_reg_alloc_concat4_i32_i128(TCGContext *s, const TCGOp *op)
+{
+ const TCGLifeData arg_life = op->life;
+ TCGTemp *ots = arg_temp(op->args[0]);
+ int be = HOST_BIG_ENDIAN ? 0xc : 0;
+
+ assert(TCG_TARGET_REG_BITS == 32);
+ tcg_debug_assert(ots->type == TCG_TYPE_I128);
+ tcg_debug_assert(!temp_readonly(ots));
+ tcg_debug_assert(NEED_SYNC_ARG(0));
+
+ if (!ots->mem_allocated) {
+ temp_allocate_frame(s, ots);
+ }
+
+ for (int i = 0; i < 4; ++i) {
+ TCGTemp *its = arg_temp(op->args[i + 1]);
+ int ofs = ots->mem_offset + ((i * 4) ^ be);
+
+ if (its->val_type == TEMP_VAL_CONST &&
+ IS_DEAD_ARG(i + 1) &&
+ tcg_out_sti(s, TCG_TYPE_I32, its->val, ots->mem_base->reg, ofs)) {
+ continue;
+ }
+
+ temp_load(s, its, tcg_target_available_regs[TCG_TYPE_I32],
+ s->reserved_regs, 0);
+ tcg_out_st(s, TCG_TYPE_I32, its->reg, ots->mem_base->reg, ofs);
+
+ if (IS_DEAD_ARG(i + 1)) {
+ temp_dead(s, its);
+ }
+ }
+
+ ots->val_type = TEMP_VAL_MEM;
+ ots->mem_coherent = 1;
+}
+
/*
* Specialized code generation for INDEX_op_dup_vec.
*/
@@ -5009,6 +5088,12 @@ int tcg_gen_code(TCGContext *s, TranslationBlock *tb, target_ulong pc_start)
case INDEX_op_mov_vec:
tcg_reg_alloc_mov(s, op);
break;
+ case INDEX_op_extr_i128_i32:
+ tcg_reg_alloc_exrl_i128_i32(s, op);
+ break;
+ case INDEX_op_concat4_i32_i128:
+ tcg_reg_alloc_concat4_i32_i128(s, op);
+ break;
case INDEX_op_dup_vec:
tcg_reg_alloc_dup(s, op);
break;
--
2.34.1
^ permalink raw reply related [flat|nested] 43+ messages in thread