* [Qemu-devel] [PULL v2 1/4] build: Use $(CCAS) for compiling .S files
2016-07-06 16:18 [Qemu-devel] [PULL v2 0/4] tcg-ish updates Richard Henderson
@ 2016-07-06 16:18 ` Richard Henderson
2016-07-07 16:16 ` Alex Bennée
2016-07-06 16:18 ` [Qemu-devel] [PULL v2 2/4] tcg: Fix name for high-half register Richard Henderson
` (3 subsequent siblings)
4 siblings, 1 reply; 10+ messages in thread
From: Richard Henderson @ 2016-07-06 16:18 UTC (permalink / raw)
To: qemu-devel; +Cc: peter.maydell
We fail to pass to $(AS) all of the different flags that may be required
for a given set of CFLAGS. Rather than figuring out the host-specific
mapping, it's better to allow the compiler driver to do that.
However, simply using $(CC) runs afoul of clang trying to build the
option roms. C.f. 3dd46c78525a30e98c68, wherein we changed from
using $(CC) to using $(AS) in the first place.
Work around this by passing -fno-integrated-as to clang, so that we use
the external assembler, and the clang driver still passes along all of
the options that the assembler might require.
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <rth@twiddle.net>
Message-Id: <1466703558-7723-1-git-send-email-rth@twiddle.net>
---
configure | 13 ++++++++++---
rules.mak | 7 ++-----
2 files changed, 12 insertions(+), 8 deletions(-)
diff --git a/configure b/configure
index 67beb47..e9090a0 100755
--- a/configure
+++ b/configure
@@ -368,7 +368,7 @@ else
fi
ar="${AR-${cross_prefix}ar}"
-as="${AS-${cross_prefix}as}"
+ccas="${CCAS-$cc}"
cpp="${CPP-$cc -E}"
objcopy="${OBJCOPY-${cross_prefix}objcopy}"
ld="${LD-${cross_prefix}ld}"
@@ -4490,6 +4490,13 @@ if test "$fortify_source" != "no"; then
fi
fi
+#################################################
+# clang does not support the 16-bit assembly for roms
+
+if echo | $ccas -dM -E - | grep __clang__ > /dev/null 2>&1 ; then
+ ccas="$ccas -fno-integrated-as"
+fi
+
##########################################
# check if struct fsxattr is available via linux/fs.h
@@ -5508,7 +5515,7 @@ echo "CXX=$cxx" >> $config_host_mak
echo "OBJCC=$objcc" >> $config_host_mak
echo "AR=$ar" >> $config_host_mak
echo "ARFLAGS=$ARFLAGS" >> $config_host_mak
-echo "AS=$as" >> $config_host_mak
+echo "CCAS=$ccas" >> $config_host_mak
echo "CPP=$cpp" >> $config_host_mak
echo "OBJCOPY=$objcopy" >> $config_host_mak
echo "LD=$ld" >> $config_host_mak
@@ -5981,7 +5988,7 @@ for rom in seabios vgabios ; do
config_mak=roms/$rom/config.mak
echo "# Automatically generated by configure - do not modify" > $config_mak
echo "SRC_PATH=$source_path/roms/$rom" >> $config_mak
- echo "AS=$as" >> $config_mak
+ echo "CCAS=$ccas" >> $config_mak
echo "CC=$cc" >> $config_mak
echo "BCC=bcc" >> $config_mak
echo "CPP=$cpp" >> $config_mak
diff --git a/rules.mak b/rules.mak
index 72c5955..7d7d83b 100644
--- a/rules.mak
+++ b/rules.mak
@@ -68,11 +68,8 @@ LINK = $(call quiet-command, $(LINKPROG) $(QEMU_CFLAGS) $(CFLAGS) $(LDFLAGS) -o
$(call process-archive-undefs, $1) \
$(version-obj-y) $(call extract-libs,$1) $(LIBS)," LINK $(TARGET_DIR)$@")
-%.asm: %.S
- $(call quiet-command,$(CPP) $(QEMU_INCLUDES) $(QEMU_CFLAGS) $(QEMU_DGFLAGS) $(CFLAGS) -o $@ $<," CPP $(TARGET_DIR)$@")
-
-%.o: %.asm
- $(call quiet-command,$(AS) $(ASFLAGS) -o $@ $<," AS $(TARGET_DIR)$@")
+%.o: %.S
+ $(call quiet-command,$(CCAS) $(QEMU_INCLUDES) $(QEMU_CFLAGS) $(QEMU_DGFLAGS) $(CFLAGS) -c -o $@ $<," AS $(TARGET_DIR)$@")
%.o: %.cc
$(call quiet-command,$(CXX) $(QEMU_INCLUDES) $(QEMU_CXXFLAGS) $(QEMU_DGFLAGS) $(CFLAGS) $($@-cflags) -c -o $@ $<," CXX $(TARGET_DIR)$@")
--
2.7.4
^ permalink raw reply related [flat|nested] 10+ messages in thread* Re: [Qemu-devel] [PULL v2 1/4] build: Use $(CCAS) for compiling .S files
2016-07-06 16:18 ` [Qemu-devel] [PULL v2 1/4] build: Use $(CCAS) for compiling .S files Richard Henderson
@ 2016-07-07 16:16 ` Alex Bennée
2016-07-07 16:29 ` Richard Henderson
2016-07-08 2:30 ` Richard Henderson
0 siblings, 2 replies; 10+ messages in thread
From: Alex Bennée @ 2016-07-07 16:16 UTC (permalink / raw)
To: Richard Henderson; +Cc: qemu-devel, peter.maydell
Richard Henderson <rth@twiddle.net> writes:
> We fail to pass to $(AS) all of the different flags that may be required
> for a given set of CFLAGS. Rather than figuring out the host-specific
> mapping, it's better to allow the compiler driver to do that.
>
> However, simply using $(CC) runs afoul of clang trying to build the
> option roms. C.f. 3dd46c78525a30e98c68, wherein we changed from
> using $(CC) to using $(AS) in the first place.
>
> Work around this by passing -fno-integrated-as to clang, so that we use
> the external assembler, and the clang driver still passes along all of
> the options that the assembler might require.
>
> Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
> Signed-off-by: Richard Henderson <rth@twiddle.net>
> Message-Id: <1466703558-7723-1-git-send-email-rth@twiddle.net>
> ---
> configure | 13 ++++++++++---
> rules.mak | 7 ++-----
> 2 files changed, 12 insertions(+), 8 deletions(-)
>
> diff --git a/configure b/configure
> index 67beb47..e9090a0 100755
> --- a/configure
> +++ b/configure
> @@ -368,7 +368,7 @@ else
> fi
>
> ar="${AR-${cross_prefix}ar}"
> -as="${AS-${cross_prefix}as}"
> +ccas="${CCAS-$cc}"
> cpp="${CPP-$cc -E}"
> objcopy="${OBJCOPY-${cross_prefix}objcopy}"
> ld="${LD-${cross_prefix}ld}"
> @@ -4490,6 +4490,13 @@ if test "$fortify_source" != "no"; then
> fi
> fi
>
> +#################################################
> +# clang does not support the 16-bit assembly for roms
> +
> +if echo | $ccas -dM -E - | grep __clang__ > /dev/null 2>&1 ; then
> + ccas="$ccas -fno-integrated-as"
> +fi
Hi Richard,
This looks like it breaks the Travis clang:
https://travis-ci.org/qemu/qemu/builds/142825178
We get the error:
AS optionrom/multiboot.o
clang: error: unknown argument: '-fno-integrated-as'
make[1]: *** [multiboot.o] Error 1
make: *** [romsubdir-optionrom] Error 2
make: *** Waiting for unfinished jobs....
Clang in Travis is:
clang version 3.4 (tags/RELEASE_34/final)
> +
> ##########################################
> # check if struct fsxattr is available via linux/fs.h
>
> @@ -5508,7 +5515,7 @@ echo "CXX=$cxx" >> $config_host_mak
> echo "OBJCC=$objcc" >> $config_host_mak
> echo "AR=$ar" >> $config_host_mak
> echo "ARFLAGS=$ARFLAGS" >> $config_host_mak
> -echo "AS=$as" >> $config_host_mak
> +echo "CCAS=$ccas" >> $config_host_mak
> echo "CPP=$cpp" >> $config_host_mak
> echo "OBJCOPY=$objcopy" >> $config_host_mak
> echo "LD=$ld" >> $config_host_mak
> @@ -5981,7 +5988,7 @@ for rom in seabios vgabios ; do
> config_mak=roms/$rom/config.mak
> echo "# Automatically generated by configure - do not modify" > $config_mak
> echo "SRC_PATH=$source_path/roms/$rom" >> $config_mak
> - echo "AS=$as" >> $config_mak
> + echo "CCAS=$ccas" >> $config_mak
> echo "CC=$cc" >> $config_mak
> echo "BCC=bcc" >> $config_mak
> echo "CPP=$cpp" >> $config_mak
> diff --git a/rules.mak b/rules.mak
> index 72c5955..7d7d83b 100644
> --- a/rules.mak
> +++ b/rules.mak
> @@ -68,11 +68,8 @@ LINK = $(call quiet-command, $(LINKPROG) $(QEMU_CFLAGS) $(CFLAGS) $(LDFLAGS) -o
> $(call process-archive-undefs, $1) \
> $(version-obj-y) $(call extract-libs,$1) $(LIBS)," LINK $(TARGET_DIR)$@")
>
> -%.asm: %.S
> - $(call quiet-command,$(CPP) $(QEMU_INCLUDES) $(QEMU_CFLAGS) $(QEMU_DGFLAGS) $(CFLAGS) -o $@ $<," CPP $(TARGET_DIR)$@")
> -
> -%.o: %.asm
> - $(call quiet-command,$(AS) $(ASFLAGS) -o $@ $<," AS $(TARGET_DIR)$@")
> +%.o: %.S
> + $(call quiet-command,$(CCAS) $(QEMU_INCLUDES) $(QEMU_CFLAGS) $(QEMU_DGFLAGS) $(CFLAGS) -c -o $@ $<," AS $(TARGET_DIR)$@")
>
> %.o: %.cc
> $(call quiet-command,$(CXX) $(QEMU_INCLUDES) $(QEMU_CXXFLAGS) $(QEMU_DGFLAGS) $(CFLAGS) $($@-cflags) -c -o $@ $<," CXX $(TARGET_DIR)$@")
--
Alex Bennée
^ permalink raw reply [flat|nested] 10+ messages in thread* Re: [Qemu-devel] [PULL v2 1/4] build: Use $(CCAS) for compiling .S files
2016-07-07 16:16 ` Alex Bennée
@ 2016-07-07 16:29 ` Richard Henderson
2016-07-07 16:32 ` Peter Maydell
2016-07-08 2:30 ` Richard Henderson
1 sibling, 1 reply; 10+ messages in thread
From: Richard Henderson @ 2016-07-07 16:29 UTC (permalink / raw)
To: Alex Bennée; +Cc: qemu-devel, peter.maydell
On 07/07/2016 09:16 AM, Alex Bennée wrote:
>> +if echo | $ccas -dM -E - | grep __clang__ > /dev/null 2>&1 ; then
>> + ccas="$ccas -fno-integrated-as"
>> +fi
>
> Hi Richard,
>
> This looks like it breaks the Travis clang:
>
> https://travis-ci.org/qemu/qemu/builds/142825178
>
> We get the error:
>
> AS optionrom/multiboot.o
> clang: error: unknown argument: '-fno-integrated-as'
> make[1]: *** [multiboot.o] Error 1
> make: *** [romsubdir-optionrom] Error 2
> make: *** Waiting for unfinished jobs....
>
> Clang in Travis is:
>
> clang version 3.4 (tags/RELEASE_34/final)
Oh ffs. It appears -fno-integrated-as was introduced in 3.5.
Is there any combination of options that allows clang 3.4 to compile the option
roms? I suspect that the answer is no. And if so... what then?
r~
^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: [Qemu-devel] [PULL v2 1/4] build: Use $(CCAS) for compiling .S files
2016-07-07 16:29 ` Richard Henderson
@ 2016-07-07 16:32 ` Peter Maydell
0 siblings, 0 replies; 10+ messages in thread
From: Peter Maydell @ 2016-07-07 16:32 UTC (permalink / raw)
To: Richard Henderson; +Cc: Alex Bennée, QEMU Developers
On 7 July 2016 at 17:29, Richard Henderson <rth@twiddle.net> wrote:
> On 07/07/2016 09:16 AM, Alex Bennée wrote:
>>>
>>> +if echo | $ccas -dM -E - | grep __clang__ > /dev/null 2>&1 ; then
>>> + ccas="$ccas -fno-integrated-as"
>>> +fi
>>
>>
>> Hi Richard,
>>
>> This looks like it breaks the Travis clang:
>>
>> https://travis-ci.org/qemu/qemu/builds/142825178
>>
>> We get the error:
>>
>> AS optionrom/multiboot.o
>> clang: error: unknown argument: '-fno-integrated-as'
>> make[1]: *** [multiboot.o] Error 1
>> make: *** [romsubdir-optionrom] Error 2
>> make: *** Waiting for unfinished jobs....
>>
>> Clang in Travis is:
>>
>> clang version 3.4 (tags/RELEASE_34/final)
>
>
> Oh ffs. It appears -fno-integrated-as was introduced in 3.5.
>
> Is there any combination of options that allows clang 3.4 to compile the
> option roms? I suspect that the answer is no. And if so... what then?
...then don't build the option roms? This would just put systems
with older clang in the same boat as non-i386 systems.
thanks
-- PMM
^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: [Qemu-devel] [PULL v2 1/4] build: Use $(CCAS) for compiling .S files
2016-07-07 16:16 ` Alex Bennée
2016-07-07 16:29 ` Richard Henderson
@ 2016-07-08 2:30 ` Richard Henderson
1 sibling, 0 replies; 10+ messages in thread
From: Richard Henderson @ 2016-07-08 2:30 UTC (permalink / raw)
To: Alex Bennée; +Cc: qemu-devel, peter.maydell
[-- Attachment #1: Type: text/plain, Size: 662 bytes --]
On 07/07/2016 09:16 AM, Alex Bennée wrote:
>> +if echo | $ccas -dM -E - | grep __clang__ > /dev/null 2>&1 ; then
>> + ccas="$ccas -fno-integrated-as"
>> +fi
>
> Hi Richard,
>
> This looks like it breaks the Travis clang:
>
> https://travis-ci.org/qemu/qemu/builds/142825178
>
> We get the error:
>
> AS optionrom/multiboot.o
> clang: error: unknown argument: '-fno-integrated-as'
> make[1]: *** [multiboot.o] Error 1
> make: *** [romsubdir-optionrom] Error 2
> make: *** Waiting for unfinished jobs....
>
> Clang in Travis is:
>
> clang version 3.4 (tags/RELEASE_34/final)
Please try this. It works here for clang 3.9.
r~
[-- Attachment #2: z --]
[-- Type: text/plain, Size: 2641 bytes --]
diff --git a/configure b/configure
index e9090a0..5ada56d 100755
--- a/configure
+++ b/configure
@@ -368,6 +368,7 @@ else
fi
ar="${AR-${cross_prefix}ar}"
+as="${AS-${cross_prefix}as}"
ccas="${CCAS-$cc}"
cpp="${CPP-$cc -E}"
objcopy="${OBJCOPY-${cross_prefix}objcopy}"
@@ -4490,13 +4491,6 @@ if test "$fortify_source" != "no"; then
fi
fi
-#################################################
-# clang does not support the 16-bit assembly for roms
-
-if echo | $ccas -dM -E - | grep __clang__ > /dev/null 2>&1 ; then
- ccas="$ccas -fno-integrated-as"
-fi
-
##########################################
# check if struct fsxattr is available via linux/fs.h
@@ -5515,6 +5509,7 @@ echo "CXX=$cxx" >> $config_host_mak
echo "OBJCC=$objcc" >> $config_host_mak
echo "AR=$ar" >> $config_host_mak
echo "ARFLAGS=$ARFLAGS" >> $config_host_mak
+echo "AS=$as" >> $config_host_mak
echo "CCAS=$ccas" >> $config_host_mak
echo "CPP=$cpp" >> $config_host_mak
echo "OBJCOPY=$objcopy" >> $config_host_mak
@@ -5988,6 +5983,7 @@ for rom in seabios vgabios ; do
config_mak=roms/$rom/config.mak
echo "# Automatically generated by configure - do not modify" > $config_mak
echo "SRC_PATH=$source_path/roms/$rom" >> $config_mak
+ echo "AS=$as" >> $config_mak
echo "CCAS=$ccas" >> $config_mak
echo "CC=$cc" >> $config_mak
echo "BCC=bcc" >> $config_mak
diff --git a/pc-bios/optionrom/Makefile b/pc-bios/optionrom/Makefile
index ce4852a..2cdda87 100644
--- a/pc-bios/optionrom/Makefile
+++ b/pc-bios/optionrom/Makefile
@@ -20,6 +20,9 @@ build-all: multiboot.bin linuxboot.bin kvmvapic.bin
# suppress auto-removal of intermediate files
.SECONDARY:
+%.o: %.S
+ $(call quiet-command,$(CPP) $(QEMU_INCLUDES) $(QEMU_DGFLAGS) $(CFLAGS) -c -o - $< | $(AS) $(ASFLAGS) -o $@," AS $(TARGET_DIR)$@")
+
%.img: %.o
$(call quiet-command,$(LD) $(LDFLAGS_NOPIE) -Ttext 0 -e _start -s -o $@ $<," Building $(TARGET_DIR)$@")
diff --git a/rules.mak b/rules.mak
index 7d7d83b..ed8e482 100644
--- a/rules.mak
+++ b/rules.mak
@@ -69,7 +69,7 @@ LINK = $(call quiet-command, $(LINKPROG) $(QEMU_CFLAGS) $(CFLAGS) $(LDFLAGS) -o
$(version-obj-y) $(call extract-libs,$1) $(LIBS)," LINK $(TARGET_DIR)$@")
%.o: %.S
- $(call quiet-command,$(CCAS) $(QEMU_INCLUDES) $(QEMU_CFLAGS) $(QEMU_DGFLAGS) $(CFLAGS) -c -o $@ $<," AS $(TARGET_DIR)$@")
+ $(call quiet-command,$(CCAS) $(QEMU_INCLUDES) $(QEMU_CFLAGS) $(QEMU_DGFLAGS) $(CFLAGS) -c -o $@ $<," CCAS $(TARGET_DIR)$@")
%.o: %.cc
$(call quiet-command,$(CXX) $(QEMU_INCLUDES) $(QEMU_CXXFLAGS) $(QEMU_DGFLAGS) $(CFLAGS) $($@-cflags) -c -o $@ $<," CXX $(TARGET_DIR)$@")
^ permalink raw reply related [flat|nested] 10+ messages in thread
* [Qemu-devel] [PULL v2 2/4] tcg: Fix name for high-half register
2016-07-06 16:18 [Qemu-devel] [PULL v2 0/4] tcg-ish updates Richard Henderson
2016-07-06 16:18 ` [Qemu-devel] [PULL v2 1/4] build: Use $(CCAS) for compiling .S files Richard Henderson
@ 2016-07-06 16:18 ` Richard Henderson
2016-07-06 16:18 ` [Qemu-devel] [PULL v2 3/4] tcg: Optimize spills of constants Richard Henderson
` (2 subsequent siblings)
4 siblings, 0 replies; 10+ messages in thread
From: Richard Henderson @ 2016-07-06 16:18 UTC (permalink / raw)
To: qemu-devel; +Cc: peter.maydell
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Richard Henderson <rth@twiddle.net>
---
tcg/tcg.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/tcg/tcg.c b/tcg/tcg.c
index 254427b..154ffe8 100644
--- a/tcg/tcg.c
+++ b/tcg/tcg.c
@@ -557,7 +557,7 @@ int tcg_global_mem_new_internal(TCGType type, TCGv_ptr base,
ts2->mem_offset = offset + (1 - bigendian) * 4;
pstrcpy(buf, sizeof(buf), name);
pstrcat(buf, sizeof(buf), "_1");
- ts->name = strdup(buf);
+ ts2->name = strdup(buf);
} else {
ts->base_type = type;
ts->type = type;
--
2.7.4
^ permalink raw reply related [flat|nested] 10+ messages in thread* [Qemu-devel] [PULL v2 3/4] tcg: Optimize spills of constants
2016-07-06 16:18 [Qemu-devel] [PULL v2 0/4] tcg-ish updates Richard Henderson
2016-07-06 16:18 ` [Qemu-devel] [PULL v2 1/4] build: Use $(CCAS) for compiling .S files Richard Henderson
2016-07-06 16:18 ` [Qemu-devel] [PULL v2 2/4] tcg: Fix name for high-half register Richard Henderson
@ 2016-07-06 16:18 ` Richard Henderson
2016-07-06 16:18 ` [Qemu-devel] [PULL v2 4/4] tcg: Improve the alignment check infrastructure Richard Henderson
2016-07-07 9:28 ` [Qemu-devel] [PULL v2 0/4] tcg-ish updates Peter Maydell
4 siblings, 0 replies; 10+ messages in thread
From: Richard Henderson @ 2016-07-06 16:18 UTC (permalink / raw)
To: qemu-devel; +Cc: peter.maydell
While we can store constants via constrants on INDEX_op_st_i32 et al,
we weren't able to spill constants to backing store.
Add a new backend interface, tcg_out_sti, which may store the constant
(and is allowed to fail). Rearrange the temp_* helpers so that we only
attempt to directly store a constant when the temp is becoming dead/free.
Signed-off-by: Richard Henderson <rth@twiddle.net>
---
tcg/aarch64/tcg-target.inc.c | 10 +++
tcg/arm/tcg-target.inc.c | 6 ++
tcg/i386/tcg-target.inc.c | 21 ++++--
tcg/ia64/tcg-target.inc.c | 10 +++
tcg/mips/tcg-target.inc.c | 10 +++
tcg/ppc/tcg-target.inc.c | 6 ++
tcg/s390/tcg-target.inc.c | 6 ++
tcg/sparc/tcg-target.inc.c | 10 +++
tcg/tcg.c | 165 ++++++++++++++++++++++++-------------------
tcg/tci/tcg-target.inc.c | 6 ++
10 files changed, 169 insertions(+), 81 deletions(-)
diff --git a/tcg/aarch64/tcg-target.inc.c b/tcg/aarch64/tcg-target.inc.c
index 1447f7c..5ac0091 100644
--- a/tcg/aarch64/tcg-target.inc.c
+++ b/tcg/aarch64/tcg-target.inc.c
@@ -716,6 +716,16 @@ static inline void tcg_out_st(TCGContext *s, TCGType type, TCGReg arg,
arg, arg1, arg2);
}
+static inline bool tcg_out_sti(TCGContext *s, TCGType type, TCGArg val,
+ TCGReg base, intptr_t ofs)
+{
+ if (val == 0) {
+ tcg_out_st(s, type, TCG_REG_XZR, base, ofs);
+ return true;
+ }
+ return false;
+}
+
static inline void tcg_out_bfm(TCGContext *s, TCGType ext, TCGReg rd,
TCGReg rn, unsigned int a, unsigned int b)
{
diff --git a/tcg/arm/tcg-target.inc.c b/tcg/arm/tcg-target.inc.c
index f9f54c6..172feba 100644
--- a/tcg/arm/tcg-target.inc.c
+++ b/tcg/arm/tcg-target.inc.c
@@ -2046,6 +2046,12 @@ static inline void tcg_out_st(TCGContext *s, TCGType type, TCGReg arg,
tcg_out_st32(s, COND_AL, arg, arg1, arg2);
}
+static inline bool tcg_out_sti(TCGContext *s, TCGType type, TCGArg val,
+ TCGReg base, intptr_t ofs)
+{
+ return false;
+}
+
static inline void tcg_out_mov(TCGContext *s, TCGType type,
TCGReg ret, TCGReg arg)
{
diff --git a/tcg/i386/tcg-target.inc.c b/tcg/i386/tcg-target.inc.c
index 317484c..bc34535 100644
--- a/tcg/i386/tcg-target.inc.c
+++ b/tcg/i386/tcg-target.inc.c
@@ -710,12 +710,19 @@ static inline void tcg_out_st(TCGContext *s, TCGType type, TCGReg arg,
tcg_out_modrm_offset(s, opc, arg, arg1, arg2);
}
-static inline void tcg_out_sti(TCGContext *s, TCGType type, TCGReg base,
- tcg_target_long ofs, tcg_target_long val)
+static bool tcg_out_sti(TCGContext *s, TCGType type, TCGArg val,
+ TCGReg base, intptr_t ofs)
{
- int opc = OPC_MOVL_EvIz + (type == TCG_TYPE_I64 ? P_REXW : 0);
- tcg_out_modrm_offset(s, opc, 0, base, ofs);
+ int rexw = 0;
+ if (TCG_TARGET_REG_BITS == 64 && type == TCG_TYPE_I64) {
+ if (val != (int32_t)val) {
+ return false;
+ }
+ rexw = P_REXW;
+ }
+ tcg_out_modrm_offset(s, OPC_MOVL_EvIz | rexw, 0, base, ofs);
tcg_out32(s, val);
+ return true;
}
static void tcg_out_shifti(TCGContext *s, int subopc, int reg, int count)
@@ -1321,10 +1328,10 @@ static void tcg_out_qemu_ld_slow_path(TCGContext *s, TCGLabelQemuLdst *l)
ofs += 4;
}
- tcg_out_sti(s, TCG_TYPE_I32, TCG_REG_ESP, ofs, oi);
+ tcg_out_sti(s, TCG_TYPE_I32, oi, TCG_REG_ESP, ofs);
ofs += 4;
- tcg_out_sti(s, TCG_TYPE_PTR, TCG_REG_ESP, ofs, (uintptr_t)l->raddr);
+ tcg_out_sti(s, TCG_TYPE_PTR, (uintptr_t)l->raddr, TCG_REG_ESP, ofs);
} else {
tcg_out_mov(s, TCG_TYPE_PTR, tcg_target_call_iarg_regs[0], TCG_AREG0);
/* The second argument is already loaded with addrlo. */
@@ -1413,7 +1420,7 @@ static void tcg_out_qemu_st_slow_path(TCGContext *s, TCGLabelQemuLdst *l)
ofs += 4;
}
- tcg_out_sti(s, TCG_TYPE_I32, TCG_REG_ESP, ofs, oi);
+ tcg_out_sti(s, TCG_TYPE_I32, oi, TCG_REG_ESP, ofs);
ofs += 4;
retaddr = TCG_REG_EAX;
diff --git a/tcg/ia64/tcg-target.inc.c b/tcg/ia64/tcg-target.inc.c
index 395223e..c91f392 100644
--- a/tcg/ia64/tcg-target.inc.c
+++ b/tcg/ia64/tcg-target.inc.c
@@ -973,6 +973,16 @@ static inline void tcg_out_st(TCGContext *s, TCGType type, TCGReg arg,
}
}
+static inline bool tcg_out_sti(TCGContext *s, TCGType type, TCGArg val,
+ TCGReg base, intptr_t ofs)
+{
+ if (val == 0) {
+ tcg_out_st(s, type, TCG_REG_R0, base, ofs);
+ return true;
+ }
+ return false;
+}
+
static inline void tcg_out_alu(TCGContext *s, uint64_t opc_a1, uint64_t opc_a3,
TCGReg ret, TCGArg arg1, int const_arg1,
TCGArg arg2, int const_arg2)
diff --git a/tcg/mips/tcg-target.inc.c b/tcg/mips/tcg-target.inc.c
index 50e98ea..2f9be48 100644
--- a/tcg/mips/tcg-target.inc.c
+++ b/tcg/mips/tcg-target.inc.c
@@ -576,6 +576,16 @@ static inline void tcg_out_st(TCGContext *s, TCGType type, TCGReg arg,
tcg_out_ldst(s, OPC_SW, arg, arg1, arg2);
}
+static inline bool tcg_out_sti(TCGContext *s, TCGType type, TCGArg val,
+ TCGReg base, intptr_t ofs)
+{
+ if (val == 0) {
+ tcg_out_st(s, type, TCG_REG_ZERO, base, ofs);
+ return true;
+ }
+ return false;
+}
+
static inline void tcg_out_addi(TCGContext *s, TCGReg reg, TCGArg val)
{
if (val == (int16_t)val) {
diff --git a/tcg/ppc/tcg-target.inc.c b/tcg/ppc/tcg-target.inc.c
index da10052..dba954c 100644
--- a/tcg/ppc/tcg-target.inc.c
+++ b/tcg/ppc/tcg-target.inc.c
@@ -857,6 +857,12 @@ static inline void tcg_out_st(TCGContext *s, TCGType type, TCGReg arg,
tcg_out_mem_long(s, opi, opx, arg, arg1, arg2);
}
+static inline bool tcg_out_sti(TCGContext *s, TCGType type, TCGArg val,
+ TCGReg base, intptr_t ofs)
+{
+ return false;
+}
+
static void tcg_out_cmp(TCGContext *s, int cond, TCGArg arg1, TCGArg arg2,
int const_arg2, int cr, TCGType type)
{
diff --git a/tcg/s390/tcg-target.inc.c b/tcg/s390/tcg-target.inc.c
index e0a60e6..314c183 100644
--- a/tcg/s390/tcg-target.inc.c
+++ b/tcg/s390/tcg-target.inc.c
@@ -798,6 +798,12 @@ static inline void tcg_out_st(TCGContext *s, TCGType type, TCGReg data,
}
}
+static inline bool tcg_out_sti(TCGContext *s, TCGType type, TCGArg val,
+ TCGReg base, intptr_t ofs)
+{
+ return false;
+}
+
/* load data from an absolute host address */
static void tcg_out_ld_abs(TCGContext *s, TCGType type, TCGReg dest, void *abs)
{
diff --git a/tcg/sparc/tcg-target.inc.c b/tcg/sparc/tcg-target.inc.c
index 9938a50..8e98172 100644
--- a/tcg/sparc/tcg-target.inc.c
+++ b/tcg/sparc/tcg-target.inc.c
@@ -504,6 +504,16 @@ static inline void tcg_out_st(TCGContext *s, TCGType type, TCGReg arg,
tcg_out_ldst(s, arg, arg1, arg2, (type == TCG_TYPE_I32 ? STW : STX));
}
+static inline bool tcg_out_sti(TCGContext *s, TCGType type, TCGArg val,
+ TCGReg base, intptr_t ofs)
+{
+ if (val == 0) {
+ tcg_out_st(s, type, TCG_REG_G0, base, ofs);
+ return true;
+ }
+ return false;
+}
+
static void tcg_out_ld_ptr(TCGContext *s, TCGReg ret, uintptr_t arg)
{
tcg_out_movi(s, TCG_TYPE_PTR, ret, arg & ~0x3ff);
diff --git a/tcg/tcg.c b/tcg/tcg.c
index 154ffe8..ef8a855 100644
--- a/tcg/tcg.c
+++ b/tcg/tcg.c
@@ -108,6 +108,8 @@ static void tcg_out_op(TCGContext *s, TCGOpcode opc, const TCGArg *args,
const int *const_args);
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);
static void tcg_out_call(TCGContext *s, tcg_insn_unit *target);
static int tcg_target_const_match(tcg_target_long val, TCGType type,
const TCGArgConstraint *arg_ct);
@@ -1680,35 +1682,89 @@ static void temp_allocate_frame(TCGContext *s, int temp)
static void temp_load(TCGContext *, TCGTemp *, TCGRegSet, TCGRegSet);
-/* sync register 'reg' by saving it to the corresponding temporary */
-static void tcg_reg_sync(TCGContext *s, TCGReg reg, TCGRegSet allocated_regs)
+/* Mark a temporary as free or dead. If 'free_or_dead' is negative,
+ mark it free; otherwise mark it dead. */
+static void temp_free_or_dead(TCGContext *s, TCGTemp *ts, int free_or_dead)
{
- TCGTemp *ts = s->reg_to_temp[reg];
+ if (ts->fixed_reg) {
+ return;
+ }
+ if (ts->val_type == TEMP_VAL_REG) {
+ s->reg_to_temp[ts->reg] = NULL;
+ }
+ ts->val_type = (free_or_dead < 0
+ || ts->temp_local
+ || temp_idx(s, ts) < s->nb_globals
+ ? TEMP_VAL_MEM : TEMP_VAL_DEAD);
+}
- tcg_debug_assert(ts->val_type == TEMP_VAL_REG);
- if (!ts->mem_coherent && !ts->fixed_reg) {
+/* Mark a temporary as dead. */
+static inline void temp_dead(TCGContext *s, TCGTemp *ts)
+{
+ temp_free_or_dead(s, ts, 1);
+}
+
+/* Sync a temporary to memory. 'allocated_regs' is used in case a temporary
+ registers needs to be allocated to store a constant. If 'free_or_dead'
+ is non-zero, subsequently release the temporary; if it is positive, the
+ temp is dead; if it is negative, the temp is free. */
+static void temp_sync(TCGContext *s, TCGTemp *ts,
+ TCGRegSet allocated_regs, int free_or_dead)
+{
+ if (ts->fixed_reg) {
+ return;
+ }
+ if (!ts->mem_coherent) {
if (!ts->mem_allocated) {
temp_allocate_frame(s, temp_idx(s, ts));
- } else if (ts->indirect_reg) {
- tcg_regset_set_reg(allocated_regs, ts->reg);
+ }
+ if (ts->indirect_reg) {
+ if (ts->val_type == TEMP_VAL_REG) {
+ tcg_regset_set_reg(allocated_regs, ts->reg);
+ }
temp_load(s, ts->mem_base,
tcg_target_available_regs[TCG_TYPE_PTR],
allocated_regs);
}
- tcg_out_st(s, ts->type, reg, ts->mem_base->reg, ts->mem_offset);
+ switch (ts->val_type) {
+ case TEMP_VAL_CONST:
+ /* If we're going to free the temp immediately, then we won't
+ require it later in a register, so attempt to store the
+ constant to memory directly. */
+ if (free_or_dead
+ && tcg_out_sti(s, ts->type, ts->val,
+ ts->mem_base->reg, ts->mem_offset)) {
+ break;
+ }
+ temp_load(s, ts, tcg_target_available_regs[ts->type],
+ allocated_regs);
+ /* fallthrough */
+
+ case TEMP_VAL_REG:
+ tcg_out_st(s, ts->type, ts->reg,
+ ts->mem_base->reg, ts->mem_offset);
+ break;
+
+ case TEMP_VAL_MEM:
+ break;
+
+ case TEMP_VAL_DEAD:
+ default:
+ tcg_abort();
+ }
+ ts->mem_coherent = 1;
+ }
+ if (free_or_dead) {
+ temp_free_or_dead(s, ts, free_or_dead);
}
- ts->mem_coherent = 1;
}
/* free register 'reg' by spilling the corresponding temporary if necessary */
static void tcg_reg_free(TCGContext *s, TCGReg reg, TCGRegSet allocated_regs)
{
TCGTemp *ts = s->reg_to_temp[reg];
-
if (ts != NULL) {
- tcg_reg_sync(s, reg, allocated_regs);
- ts->val_type = TEMP_VAL_MEM;
- s->reg_to_temp[reg] = NULL;
+ temp_sync(s, ts, allocated_regs, -1);
}
}
@@ -1778,45 +1834,9 @@ static void temp_load(TCGContext *s, TCGTemp *ts, TCGRegSet desired_regs,
s->reg_to_temp[reg] = ts;
}
-/* mark a temporary as dead. */
-static inline void temp_dead(TCGContext *s, TCGTemp *ts)
-{
- if (ts->fixed_reg) {
- return;
- }
- if (ts->val_type == TEMP_VAL_REG) {
- s->reg_to_temp[ts->reg] = NULL;
- }
- ts->val_type = (temp_idx(s, ts) < s->nb_globals || ts->temp_local
- ? TEMP_VAL_MEM : TEMP_VAL_DEAD);
-}
-
-/* sync a temporary to memory. 'allocated_regs' is used in case a
- temporary registers needs to be allocated to store a constant. */
-static void temp_sync(TCGContext *s, TCGTemp *ts, TCGRegSet allocated_regs)
-{
- if (ts->fixed_reg) {
- return;
- }
- switch (ts->val_type) {
- case TEMP_VAL_CONST:
- temp_load(s, ts, tcg_target_available_regs[ts->type], allocated_regs);
- /* fallthrough */
- case TEMP_VAL_REG:
- tcg_reg_sync(s, ts->reg, allocated_regs);
- break;
- case TEMP_VAL_DEAD:
- case TEMP_VAL_MEM:
- break;
- default:
- tcg_abort();
- }
-}
-
-/* save a temporary to memory. 'allocated_regs' is used in case a
- temporary registers needs to be allocated to store a constant. */
-static inline void temp_save(TCGContext *s, TCGTemp *ts,
- TCGRegSet allocated_regs)
+/* Save a temporary to memory. 'allocated_regs' is used in case a
+ temporary registers needs to be allocated to store a constant. */
+static void temp_save(TCGContext *s, TCGTemp *ts, TCGRegSet allocated_regs)
{
#ifdef USE_LIVENESS_ANALYSIS
/* ??? Liveness does not yet incorporate indirect bases. */
@@ -1827,8 +1847,7 @@ static inline void temp_save(TCGContext *s, TCGTemp *ts,
return;
}
#endif
- temp_sync(s, ts, allocated_regs);
- temp_dead(s, ts);
+ temp_sync(s, ts, allocated_regs, 1);
}
/* save globals to their canonical location and assume they can be
@@ -1861,7 +1880,7 @@ static void sync_globals(TCGContext *s, TCGRegSet allocated_regs)
continue;
}
#endif
- temp_sync(s, ts, allocated_regs);
+ temp_sync(s, ts, allocated_regs, 0);
}
}
@@ -1905,21 +1924,21 @@ static void tcg_reg_alloc_movi(TCGContext *s, const TCGArg *args,
val = args[1];
if (ots->fixed_reg) {
- /* for fixed registers, we do not do any constant
- propagation */
+ /* For fixed registers, we do not do any constant propagation. */
tcg_out_movi(s, ots->type, ots->reg, val);
- } else {
- /* The movi is not explicitly generated here */
- if (ots->val_type == TEMP_VAL_REG) {
- s->reg_to_temp[ots->reg] = NULL;
- }
- ots->val_type = TEMP_VAL_CONST;
- ots->val = val;
+ return;
}
- if (NEED_SYNC_ARG(0)) {
- temp_sync(s, ots, s->reserved_regs);
+
+ /* The movi is not explicitly generated here. */
+ if (ots->val_type == TEMP_VAL_REG) {
+ s->reg_to_temp[ots->reg] = NULL;
}
- if (IS_DEAD_ARG(0)) {
+ ots->val_type = TEMP_VAL_CONST;
+ ots->val = val;
+ ots->mem_coherent = 0;
+ if (NEED_SYNC_ARG(0)) {
+ temp_sync(s, ots, s->reserved_regs, IS_DEAD_ARG(0));
+ } else if (IS_DEAD_ARG(0)) {
temp_dead(s, ots);
}
}
@@ -2004,7 +2023,7 @@ static void tcg_reg_alloc_mov(TCGContext *s, const TCGOpDef *def,
ots->mem_coherent = 0;
s->reg_to_temp[ots->reg] = ots;
if (NEED_SYNC_ARG(0)) {
- tcg_reg_sync(s, ots->reg, allocated_regs);
+ temp_sync(s, ots, allocated_regs, 0);
}
}
}
@@ -2163,9 +2182,8 @@ static void tcg_reg_alloc_op(TCGContext *s,
tcg_out_mov(s, ts->type, ts->reg, reg);
}
if (NEED_SYNC_ARG(i)) {
- tcg_reg_sync(s, reg, allocated_regs);
- }
- if (IS_DEAD_ARG(i)) {
+ temp_sync(s, ts, allocated_regs, IS_DEAD_ARG(i));
+ } else if (IS_DEAD_ARG(i)) {
temp_dead(s, ts);
}
}
@@ -2298,9 +2316,8 @@ static void tcg_reg_alloc_call(TCGContext *s, int nb_oargs, int nb_iargs,
ts->mem_coherent = 0;
s->reg_to_temp[reg] = ts;
if (NEED_SYNC_ARG(i)) {
- tcg_reg_sync(s, reg, allocated_regs);
- }
- if (IS_DEAD_ARG(i)) {
+ temp_sync(s, ts, allocated_regs, IS_DEAD_ARG(i));
+ } else if (IS_DEAD_ARG(i)) {
temp_dead(s, ts);
}
}
diff --git a/tcg/tci/tcg-target.inc.c b/tcg/tci/tcg-target.inc.c
index fa74d52..3c47ea7 100644
--- a/tcg/tci/tcg-target.inc.c
+++ b/tcg/tci/tcg-target.inc.c
@@ -834,6 +834,12 @@ static void tcg_out_st(TCGContext *s, TCGType type, TCGReg arg, TCGReg arg1,
old_code_ptr[1] = s->code_ptr - old_code_ptr;
}
+static inline bool tcg_out_sti(TCGContext *s, TCGType type, TCGArg val,
+ TCGReg base, intptr_t ofs)
+{
+ return false;
+}
+
/* Test if a constant matches the constraint. */
static int tcg_target_const_match(tcg_target_long val, TCGType type,
const TCGArgConstraint *arg_ct)
--
2.7.4
^ permalink raw reply related [flat|nested] 10+ messages in thread* [Qemu-devel] [PULL v2 4/4] tcg: Improve the alignment check infrastructure
2016-07-06 16:18 [Qemu-devel] [PULL v2 0/4] tcg-ish updates Richard Henderson
` (2 preceding siblings ...)
2016-07-06 16:18 ` [Qemu-devel] [PULL v2 3/4] tcg: Optimize spills of constants Richard Henderson
@ 2016-07-06 16:18 ` Richard Henderson
2016-07-07 9:28 ` [Qemu-devel] [PULL v2 0/4] tcg-ish updates Peter Maydell
4 siblings, 0 replies; 10+ messages in thread
From: Richard Henderson @ 2016-07-06 16:18 UTC (permalink / raw)
To: qemu-devel; +Cc: peter.maydell, Sergey Sorokin
From: Sergey Sorokin <afarallax@yandex.ru>
Some architectures (e.g. ARMv8) need the address which is aligned
to a size more than the size of the memory access.
To support such check it's enough the current costless alignment
check implementation in QEMU, but we need to support
an alignment size specifying.
Signed-off-by: Sergey Sorokin <afarallax@yandex.ru>
Message-Id: <1466705806-679898-1-git-send-email-afarallax@yandex.ru>
Signed-off-by: Richard Henderson <rth@twiddle.net>
[rth: Assert in tcg_canonicalize_memop. Leave get_alignment_bits
available for, though unused by, user-mode. Retain logging difference
based on ALIGNED_ONLY.]
---
include/exec/cpu-all.h | 16 ++++++--
softmmu_template.h | 88 ++++++++++++--------------------------------
tcg/aarch64/tcg-target.inc.c | 9 +++--
tcg/i386/tcg-target.inc.c | 15 +++++---
tcg/ppc/tcg-target.inc.c | 14 ++++---
tcg/s390/tcg-target.inc.c | 9 +++--
tcg/tcg-op.c | 3 ++
tcg/tcg.c | 26 +++++++++----
tcg/tcg.h | 87 ++++++++++++++++++++++++++++++++++++-------
9 files changed, 159 insertions(+), 108 deletions(-)
diff --git a/include/exec/cpu-all.h b/include/exec/cpu-all.h
index 6a6796d..b6a7059 100644
--- a/include/exec/cpu-all.h
+++ b/include/exec/cpu-all.h
@@ -288,14 +288,22 @@ CPUArchState *cpu_copy(CPUArchState *env);
#if !defined(CONFIG_USER_ONLY)
/* Flags stored in the low bits of the TLB virtual address. These are
- defined so that fast path ram access is all zeros. */
+ * defined so that fast path ram access is all zeros.
+ * The flags all must be between TARGET_PAGE_BITS and
+ * maximum address alignment bit.
+ */
/* Zero if TLB entry is valid. */
-#define TLB_INVALID_MASK (1 << 3)
+#define TLB_INVALID_MASK (1 << (TARGET_PAGE_BITS - 1))
/* Set if TLB entry references a clean RAM page. The iotlb entry will
contain the page physical address. */
-#define TLB_NOTDIRTY (1 << 4)
+#define TLB_NOTDIRTY (1 << (TARGET_PAGE_BITS - 2))
/* Set if TLB entry is an IO callback. */
-#define TLB_MMIO (1 << 5)
+#define TLB_MMIO (1 << (TARGET_PAGE_BITS - 3))
+
+/* Use this mask to check interception with an alignment mask
+ * in a TCG backend.
+ */
+#define TLB_FLAGS_MASK (TLB_INVALID_MASK | TLB_NOTDIRTY | TLB_MMIO)
void dump_exec_info(FILE *f, fprintf_function cpu_fprintf);
void dump_opcount_info(FILE *f, fprintf_function cpu_fprintf);
diff --git a/softmmu_template.h b/softmmu_template.h
index 208f808..4d378ca 100644
--- a/softmmu_template.h
+++ b/softmmu_template.h
@@ -171,20 +171,21 @@ WORD_TYPE helper_le_ld_name(CPUArchState *env, target_ulong addr,
unsigned mmu_idx = get_mmuidx(oi);
int index = (addr >> TARGET_PAGE_BITS) & (CPU_TLB_SIZE - 1);
target_ulong tlb_addr = env->tlb_table[mmu_idx][index].ADDR_READ;
+ int a_bits = get_alignment_bits(get_memop(oi));
uintptr_t haddr;
DATA_TYPE res;
/* Adjust the given return address. */
retaddr -= GETPC_ADJ;
+ if (a_bits > 0 && (addr & ((1 << a_bits) - 1)) != 0) {
+ cpu_unaligned_access(ENV_GET_CPU(env), addr, READ_ACCESS_TYPE,
+ mmu_idx, retaddr);
+ }
+
/* If the TLB entry is for a different page, reload and try again. */
if ((addr & TARGET_PAGE_MASK)
!= (tlb_addr & (TARGET_PAGE_MASK | TLB_INVALID_MASK))) {
- if ((addr & (DATA_SIZE - 1)) != 0
- && (get_memop(oi) & MO_AMASK) == MO_ALIGN) {
- cpu_unaligned_access(ENV_GET_CPU(env), addr, READ_ACCESS_TYPE,
- mmu_idx, retaddr);
- }
if (!VICTIM_TLB_HIT(ADDR_READ)) {
tlb_fill(ENV_GET_CPU(env), addr, READ_ACCESS_TYPE,
mmu_idx, retaddr);
@@ -215,10 +216,6 @@ WORD_TYPE helper_le_ld_name(CPUArchState *env, target_ulong addr,
DATA_TYPE res1, res2;
unsigned shift;
do_unaligned_access:
- if ((get_memop(oi) & MO_AMASK) == MO_ALIGN) {
- cpu_unaligned_access(ENV_GET_CPU(env), addr, READ_ACCESS_TYPE,
- mmu_idx, retaddr);
- }
addr1 = addr & ~(DATA_SIZE - 1);
addr2 = addr1 + DATA_SIZE;
/* Note the adjustment at the beginning of the function.
@@ -232,13 +229,6 @@ WORD_TYPE helper_le_ld_name(CPUArchState *env, target_ulong addr,
return res;
}
- /* Handle aligned access or unaligned access in the same page. */
- if ((addr & (DATA_SIZE - 1)) != 0
- && (get_memop(oi) & MO_AMASK) == MO_ALIGN) {
- cpu_unaligned_access(ENV_GET_CPU(env), addr, READ_ACCESS_TYPE,
- mmu_idx, retaddr);
- }
-
haddr = addr + env->tlb_table[mmu_idx][index].addend;
#if DATA_SIZE == 1
res = glue(glue(ld, LSUFFIX), _p)((uint8_t *)haddr);
@@ -255,20 +245,21 @@ WORD_TYPE helper_be_ld_name(CPUArchState *env, target_ulong addr,
unsigned mmu_idx = get_mmuidx(oi);
int index = (addr >> TARGET_PAGE_BITS) & (CPU_TLB_SIZE - 1);
target_ulong tlb_addr = env->tlb_table[mmu_idx][index].ADDR_READ;
+ int a_bits = get_alignment_bits(get_memop(oi));
uintptr_t haddr;
DATA_TYPE res;
/* Adjust the given return address. */
retaddr -= GETPC_ADJ;
+ if (a_bits > 0 && (addr & ((1 << a_bits) - 1)) != 0) {
+ cpu_unaligned_access(ENV_GET_CPU(env), addr, READ_ACCESS_TYPE,
+ mmu_idx, retaddr);
+ }
+
/* If the TLB entry is for a different page, reload and try again. */
if ((addr & TARGET_PAGE_MASK)
!= (tlb_addr & (TARGET_PAGE_MASK | TLB_INVALID_MASK))) {
- if ((addr & (DATA_SIZE - 1)) != 0
- && (get_memop(oi) & MO_AMASK) == MO_ALIGN) {
- cpu_unaligned_access(ENV_GET_CPU(env), addr, READ_ACCESS_TYPE,
- mmu_idx, retaddr);
- }
if (!VICTIM_TLB_HIT(ADDR_READ)) {
tlb_fill(ENV_GET_CPU(env), addr, READ_ACCESS_TYPE,
mmu_idx, retaddr);
@@ -299,10 +290,6 @@ WORD_TYPE helper_be_ld_name(CPUArchState *env, target_ulong addr,
DATA_TYPE res1, res2;
unsigned shift;
do_unaligned_access:
- if ((get_memop(oi) & MO_AMASK) == MO_ALIGN) {
- cpu_unaligned_access(ENV_GET_CPU(env), addr, READ_ACCESS_TYPE,
- mmu_idx, retaddr);
- }
addr1 = addr & ~(DATA_SIZE - 1);
addr2 = addr1 + DATA_SIZE;
/* Note the adjustment at the beginning of the function.
@@ -316,13 +303,6 @@ WORD_TYPE helper_be_ld_name(CPUArchState *env, target_ulong addr,
return res;
}
- /* Handle aligned access or unaligned access in the same page. */
- if ((addr & (DATA_SIZE - 1)) != 0
- && (get_memop(oi) & MO_AMASK) == MO_ALIGN) {
- cpu_unaligned_access(ENV_GET_CPU(env), addr, READ_ACCESS_TYPE,
- mmu_idx, retaddr);
- }
-
haddr = addr + env->tlb_table[mmu_idx][index].addend;
res = glue(glue(ld, LSUFFIX), _be_p)((uint8_t *)haddr);
return res;
@@ -376,19 +356,20 @@ void helper_le_st_name(CPUArchState *env, target_ulong addr, DATA_TYPE val,
unsigned mmu_idx = get_mmuidx(oi);
int index = (addr >> TARGET_PAGE_BITS) & (CPU_TLB_SIZE - 1);
target_ulong tlb_addr = env->tlb_table[mmu_idx][index].addr_write;
+ int a_bits = get_alignment_bits(get_memop(oi));
uintptr_t haddr;
/* Adjust the given return address. */
retaddr -= GETPC_ADJ;
+ if (a_bits > 0 && (addr & ((1 << a_bits) - 1)) != 0) {
+ cpu_unaligned_access(ENV_GET_CPU(env), addr, MMU_DATA_STORE,
+ mmu_idx, retaddr);
+ }
+
/* If the TLB entry is for a different page, reload and try again. */
if ((addr & TARGET_PAGE_MASK)
!= (tlb_addr & (TARGET_PAGE_MASK | TLB_INVALID_MASK))) {
- if ((addr & (DATA_SIZE - 1)) != 0
- && (get_memop(oi) & MO_AMASK) == MO_ALIGN) {
- cpu_unaligned_access(ENV_GET_CPU(env), addr, MMU_DATA_STORE,
- mmu_idx, retaddr);
- }
if (!VICTIM_TLB_HIT(addr_write)) {
tlb_fill(ENV_GET_CPU(env), addr, MMU_DATA_STORE, mmu_idx, retaddr);
}
@@ -416,10 +397,6 @@ void helper_le_st_name(CPUArchState *env, target_ulong addr, DATA_TYPE val,
>= TARGET_PAGE_SIZE)) {
int i;
do_unaligned_access:
- if ((get_memop(oi) & MO_AMASK) == MO_ALIGN) {
- cpu_unaligned_access(ENV_GET_CPU(env), addr, MMU_DATA_STORE,
- mmu_idx, retaddr);
- }
/* XXX: not efficient, but simple */
/* Note: relies on the fact that tlb_fill() does not remove the
* previous page from the TLB cache. */
@@ -434,13 +411,6 @@ void helper_le_st_name(CPUArchState *env, target_ulong addr, DATA_TYPE val,
return;
}
- /* Handle aligned access or unaligned access in the same page. */
- if ((addr & (DATA_SIZE - 1)) != 0
- && (get_memop(oi) & MO_AMASK) == MO_ALIGN) {
- cpu_unaligned_access(ENV_GET_CPU(env), addr, MMU_DATA_STORE,
- mmu_idx, retaddr);
- }
-
haddr = addr + env->tlb_table[mmu_idx][index].addend;
#if DATA_SIZE == 1
glue(glue(st, SUFFIX), _p)((uint8_t *)haddr, val);
@@ -456,19 +426,20 @@ void helper_be_st_name(CPUArchState *env, target_ulong addr, DATA_TYPE val,
unsigned mmu_idx = get_mmuidx(oi);
int index = (addr >> TARGET_PAGE_BITS) & (CPU_TLB_SIZE - 1);
target_ulong tlb_addr = env->tlb_table[mmu_idx][index].addr_write;
+ int a_bits = get_alignment_bits(get_memop(oi));
uintptr_t haddr;
/* Adjust the given return address. */
retaddr -= GETPC_ADJ;
+ if (a_bits > 0 && (addr & ((1 << a_bits) - 1)) != 0) {
+ cpu_unaligned_access(ENV_GET_CPU(env), addr, MMU_DATA_STORE,
+ mmu_idx, retaddr);
+ }
+
/* If the TLB entry is for a different page, reload and try again. */
if ((addr & TARGET_PAGE_MASK)
!= (tlb_addr & (TARGET_PAGE_MASK | TLB_INVALID_MASK))) {
- if ((addr & (DATA_SIZE - 1)) != 0
- && (get_memop(oi) & MO_AMASK) == MO_ALIGN) {
- cpu_unaligned_access(ENV_GET_CPU(env), addr, MMU_DATA_STORE,
- mmu_idx, retaddr);
- }
if (!VICTIM_TLB_HIT(addr_write)) {
tlb_fill(ENV_GET_CPU(env), addr, MMU_DATA_STORE, mmu_idx, retaddr);
}
@@ -496,10 +467,6 @@ void helper_be_st_name(CPUArchState *env, target_ulong addr, DATA_TYPE val,
>= TARGET_PAGE_SIZE)) {
int i;
do_unaligned_access:
- if ((get_memop(oi) & MO_AMASK) == MO_ALIGN) {
- cpu_unaligned_access(ENV_GET_CPU(env), addr, MMU_DATA_STORE,
- mmu_idx, retaddr);
- }
/* XXX: not efficient, but simple */
/* Note: relies on the fact that tlb_fill() does not remove the
* previous page from the TLB cache. */
@@ -514,13 +481,6 @@ void helper_be_st_name(CPUArchState *env, target_ulong addr, DATA_TYPE val,
return;
}
- /* Handle aligned access or unaligned access in the same page. */
- if ((addr & (DATA_SIZE - 1)) != 0
- && (get_memop(oi) & MO_AMASK) == MO_ALIGN) {
- cpu_unaligned_access(ENV_GET_CPU(env), addr, MMU_DATA_STORE,
- mmu_idx, retaddr);
- }
-
haddr = addr + env->tlb_table[mmu_idx][index].addend;
glue(glue(st, SUFFIX), _be_p)((uint8_t *)haddr, val);
}
diff --git a/tcg/aarch64/tcg-target.inc.c b/tcg/aarch64/tcg-target.inc.c
index 5ac0091..08b2d03 100644
--- a/tcg/aarch64/tcg-target.inc.c
+++ b/tcg/aarch64/tcg-target.inc.c
@@ -1081,19 +1081,20 @@ static void tcg_out_tlb_read(TCGContext *s, TCGReg addr_reg, TCGMemOp opc,
int tlb_offset = is_read ?
offsetof(CPUArchState, tlb_table[mem_index][0].addr_read)
: offsetof(CPUArchState, tlb_table[mem_index][0].addr_write);
- int s_mask = (1 << (opc & MO_SIZE)) - 1;
+ int a_bits = get_alignment_bits(opc);
TCGReg base = TCG_AREG0, x3;
uint64_t tlb_mask;
/* For aligned accesses, we check the first byte and include the alignment
bits within the address. For unaligned access, we check that we don't
cross pages using the address of the last byte of the access. */
- if ((opc & MO_AMASK) == MO_ALIGN || s_mask == 0) {
- tlb_mask = TARGET_PAGE_MASK | s_mask;
+ if (a_bits >= 0) {
+ /* A byte access or an alignment check required */
+ tlb_mask = TARGET_PAGE_MASK | ((1 << a_bits) - 1);
x3 = addr_reg;
} else {
tcg_out_insn(s, 3401, ADDI, TARGET_LONG_BITS == 64,
- TCG_REG_X3, addr_reg, s_mask);
+ TCG_REG_X3, addr_reg, (1 << (opc & MO_SIZE)) - 1);
tlb_mask = TARGET_PAGE_MASK;
x3 = TCG_REG_X3;
}
diff --git a/tcg/i386/tcg-target.inc.c b/tcg/i386/tcg-target.inc.c
index bc34535..6f8cdca 100644
--- a/tcg/i386/tcg-target.inc.c
+++ b/tcg/i386/tcg-target.inc.c
@@ -1202,8 +1202,8 @@ static inline void tcg_out_tlb_load(TCGContext *s, TCGReg addrlo, TCGReg addrhi,
TCGType ttype = TCG_TYPE_I32;
TCGType tlbtype = TCG_TYPE_I32;
int trexw = 0, hrexw = 0, tlbrexw = 0;
- int s_mask = (1 << (opc & MO_SIZE)) - 1;
- bool aligned = (opc & MO_AMASK) == MO_ALIGN || s_mask == 0;
+ int a_bits = get_alignment_bits(opc);
+ target_ulong tlb_mask;
if (TCG_TARGET_REG_BITS == 64) {
if (TARGET_LONG_BITS == 64) {
@@ -1220,19 +1220,22 @@ static inline void tcg_out_tlb_load(TCGContext *s, TCGReg addrlo, TCGReg addrhi,
}
tcg_out_mov(s, tlbtype, r0, addrlo);
- if (aligned) {
+ if (a_bits >= 0) {
+ /* A byte access or an alignment check required */
tcg_out_mov(s, ttype, r1, addrlo);
+ tlb_mask = TARGET_PAGE_MASK | ((1 << a_bits) - 1);
} else {
/* For unaligned access check that we don't cross pages using
the page address of the last byte. */
- tcg_out_modrm_offset(s, OPC_LEA + trexw, r1, addrlo, s_mask);
+ tcg_out_modrm_offset(s, OPC_LEA + trexw, r1, addrlo,
+ (1 << (opc & MO_SIZE)) - 1);
+ tlb_mask = TARGET_PAGE_MASK;
}
tcg_out_shifti(s, SHIFT_SHR + tlbrexw, r0,
TARGET_PAGE_BITS - CPU_TLB_ENTRY_BITS);
- tgen_arithi(s, ARITH_AND + trexw, r1,
- TARGET_PAGE_MASK | (aligned ? s_mask : 0), 0);
+ tgen_arithi(s, ARITH_AND + trexw, r1, tlb_mask, 0);
tgen_arithi(s, ARITH_AND + tlbrexw, r0,
(CPU_TLB_SIZE - 1) << CPU_TLB_ENTRY_BITS, 0);
diff --git a/tcg/ppc/tcg-target.inc.c b/tcg/ppc/tcg-target.inc.c
index dba954c..eaf1bd9 100644
--- a/tcg/ppc/tcg-target.inc.c
+++ b/tcg/ppc/tcg-target.inc.c
@@ -1405,6 +1405,7 @@ static TCGReg tcg_out_tlb_read(TCGContext *s, TCGMemOp opc,
int add_off = offsetof(CPUArchState, tlb_table[mem_index][0].addend);
TCGReg base = TCG_AREG0;
TCGMemOp s_bits = opc & MO_SIZE;
+ int a_bits = get_alignment_bits(opc);
/* Extract the page index, shifted into place for tlb index. */
if (TCG_TARGET_REG_BITS == 64) {
@@ -1462,14 +1463,17 @@ static TCGReg tcg_out_tlb_read(TCGContext *s, TCGMemOp opc,
* the bottom bits and thus trigger a comparison failure on
* unaligned accesses
*/
+ if (a_bits < 0) {
+ a_bits = s_bits;
+ }
tcg_out_rlw(s, RLWINM, TCG_REG_R0, addrlo, 0,
- (32 - s_bits) & 31, 31 - TARGET_PAGE_BITS);
- } else if (s_bits) {
- /* > byte access, we need to handle alignment */
- if ((opc & MO_AMASK) == MO_ALIGN) {
+ (32 - a_bits) & 31, 31 - TARGET_PAGE_BITS);
+ } else if (a_bits) {
+ /* More than byte access, we need to handle alignment */
+ if (a_bits > 0) {
/* Alignment required by the front-end, same as 32-bits */
tcg_out_rld(s, RLDICL, TCG_REG_R0, addrlo,
- 64 - TARGET_PAGE_BITS, TARGET_PAGE_BITS - s_bits);
+ 64 - TARGET_PAGE_BITS, TARGET_PAGE_BITS - a_bits);
tcg_out_rld(s, RLDICL, TCG_REG_R0, TCG_REG_R0, TARGET_PAGE_BITS, 0);
} else {
/* We support unaligned accesses, we need to make sure we fail
diff --git a/tcg/s390/tcg-target.inc.c b/tcg/s390/tcg-target.inc.c
index 314c183..5a7495b 100644
--- a/tcg/s390/tcg-target.inc.c
+++ b/tcg/s390/tcg-target.inc.c
@@ -1505,18 +1505,19 @@ QEMU_BUILD_BUG_ON(offsetof(CPUArchState, tlb_table[NB_MMU_MODES - 1][1])
static TCGReg tcg_out_tlb_read(TCGContext* s, TCGReg addr_reg, TCGMemOp opc,
int mem_index, bool is_ld)
{
- int s_mask = (1 << (opc & MO_SIZE)) - 1;
+ int a_bits = get_alignment_bits(opc);
int ofs, a_off;
uint64_t tlb_mask;
/* For aligned accesses, we check the first byte and include the alignment
bits within the address. For unaligned access, we check that we don't
cross pages using the address of the last byte of the access. */
- if ((opc & MO_AMASK) == MO_ALIGN || s_mask == 0) {
+ if (a_bits >= 0) {
+ /* A byte access or an alignment check required */
a_off = 0;
- tlb_mask = TARGET_PAGE_MASK | s_mask;
+ tlb_mask = TARGET_PAGE_MASK | ((1 << a_bits) - 1);
} else {
- a_off = s_mask;
+ a_off = (1 << (opc & MO_SIZE)) - 1;
tlb_mask = TARGET_PAGE_MASK;
}
diff --git a/tcg/tcg-op.c b/tcg/tcg-op.c
index 569cdc6..293b854 100644
--- a/tcg/tcg-op.c
+++ b/tcg/tcg-op.c
@@ -1851,6 +1851,9 @@ void tcg_gen_goto_tb(unsigned idx)
static inline TCGMemOp tcg_canonicalize_memop(TCGMemOp op, bool is64, bool st)
{
+ /* Trigger the asserts within as early as possible. */
+ (void)get_alignment_bits(op);
+
switch (op & MO_SIZE) {
case MO_8:
op &= ~MO_BSWAP;
diff --git a/tcg/tcg.c b/tcg/tcg.c
index ef8a855..0c46c43 100644
--- a/tcg/tcg.c
+++ b/tcg/tcg.c
@@ -999,6 +999,22 @@ static const char * const ldst_name[] =
[MO_BEQ] = "beq",
};
+static const char * const alignment_name[(MO_AMASK >> MO_ASHIFT) + 1] = {
+#ifdef ALIGNED_ONLY
+ [MO_UNALN >> MO_ASHIFT] = "un+",
+ [MO_ALIGN >> MO_ASHIFT] = "",
+#else
+ [MO_UNALN >> MO_ASHIFT] = "",
+ [MO_ALIGN >> MO_ASHIFT] = "al+",
+#endif
+ [MO_ALIGN_2 >> MO_ASHIFT] = "al2+",
+ [MO_ALIGN_4 >> MO_ASHIFT] = "al4+",
+ [MO_ALIGN_8 >> MO_ASHIFT] = "al8+",
+ [MO_ALIGN_16 >> MO_ASHIFT] = "al16+",
+ [MO_ALIGN_32 >> MO_ASHIFT] = "al32+",
+ [MO_ALIGN_64 >> MO_ASHIFT] = "al64+",
+};
+
void tcg_dump_ops(TCGContext *s)
{
char buf[128];
@@ -1100,14 +1116,8 @@ void tcg_dump_ops(TCGContext *s)
if (op & ~(MO_AMASK | MO_BSWAP | MO_SSIZE)) {
qemu_log(",$0x%x,%u", op, ix);
} else {
- const char *s_al = "", *s_op;
- if (op & MO_AMASK) {
- if ((op & MO_AMASK) == MO_ALIGN) {
- s_al = "al+";
- } else {
- s_al = "un+";
- }
- }
+ const char *s_al, *s_op;
+ s_al = alignment_name[(op & MO_AMASK) >> MO_ASHIFT];
s_op = ldst_name[op & (MO_BSWAP | MO_SSIZE)];
qemu_log(",%s%s,%u", s_al, s_op, ix);
}
diff --git a/tcg/tcg.h b/tcg/tcg.h
index 66d7fc0..66ae0c7 100644
--- a/tcg/tcg.h
+++ b/tcg/tcg.h
@@ -191,6 +191,15 @@ typedef uint64_t tcg_insn_unit;
#endif
+#ifdef CONFIG_DEBUG_TCG
+# define tcg_debug_assert(X) do { assert(X); } while (0)
+#elif QEMU_GNUC_PREREQ(4, 5)
+# define tcg_debug_assert(X) \
+ do { if (!(X)) { __builtin_unreachable(); } } while (0)
+#else
+# define tcg_debug_assert(X) do { (void)(X); } while (0)
+#endif
+
typedef struct TCGRelocation {
struct TCGRelocation *next;
int type;
@@ -275,10 +284,26 @@ typedef enum TCGMemOp {
#endif
/* MO_UNALN accesses are never checked for alignment.
- MO_ALIGN accesses will result in a call to the CPU's
- do_unaligned_access hook if the guest address is not aligned.
- The default depends on whether the target CPU defines ALIGNED_ONLY. */
- MO_AMASK = 16,
+ * MO_ALIGN accesses will result in a call to the CPU's
+ * do_unaligned_access hook if the guest address is not aligned.
+ * The default depends on whether the target CPU defines ALIGNED_ONLY.
+ * Some architectures (e.g. ARMv8) need the address which is aligned
+ * to a size more than the size of the memory access.
+ * To support such check it's enough the current costless alignment
+ * check implementation in QEMU, but we need to support
+ * an alignment size specifying.
+ * MO_ALIGN supposes a natural alignment
+ * (i.e. the alignment size is the size of a memory access).
+ * Note that an alignment size must be equal or greater
+ * than an access size.
+ * There are three options:
+ * - an alignment to the size of an access (MO_ALIGN);
+ * - an alignment to the specified size that is equal or greater than
+ * an access size (MO_ALIGN_x where 'x' is a size in bytes);
+ * - unaligned access permitted (MO_UNALN).
+ */
+ MO_ASHIFT = 4,
+ MO_AMASK = 7 << MO_ASHIFT,
#ifdef ALIGNED_ONLY
MO_ALIGN = 0,
MO_UNALN = MO_AMASK,
@@ -286,6 +311,12 @@ typedef enum TCGMemOp {
MO_ALIGN = MO_AMASK,
MO_UNALN = 0,
#endif
+ MO_ALIGN_2 = 1 << MO_ASHIFT,
+ MO_ALIGN_4 = 2 << MO_ASHIFT,
+ MO_ALIGN_8 = 3 << MO_ASHIFT,
+ MO_ALIGN_16 = 4 << MO_ASHIFT,
+ MO_ALIGN_32 = 5 << MO_ASHIFT,
+ MO_ALIGN_64 = 6 << MO_ASHIFT,
/* Combinations of the above, for ease of use. */
MO_UB = MO_8,
@@ -317,6 +348,45 @@ typedef enum TCGMemOp {
MO_SSIZE = MO_SIZE | MO_SIGN,
} TCGMemOp;
+/**
+ * get_alignment_bits
+ * @memop: TCGMemOp value
+ *
+ * Extract the alignment size from the memop.
+ *
+ * Returns: 0 in case of byte access (which is always aligned);
+ * positive value - number of alignment bits;
+ * negative value if unaligned access enabled
+ * and this is not a byte access.
+ */
+static inline int get_alignment_bits(TCGMemOp memop)
+{
+ int a = memop & MO_AMASK;
+ int s = memop & MO_SIZE;
+ int r;
+
+ if (a == MO_UNALN) {
+ /* Negative value if unaligned access enabled,
+ * or zero value in case of byte access.
+ */
+ return -s;
+ } else if (a == MO_ALIGN) {
+ /* A natural alignment: return a number of access size bits */
+ r = s;
+ } else {
+ /* Specific alignment size. It must be equal or greater
+ * than the access size.
+ */
+ r = a >> MO_ASHIFT;
+ tcg_debug_assert(r >= s);
+ }
+#if defined(CONFIG_SOFTMMU)
+ /* The requested alignment cannot overlap the TLB flags. */
+ tcg_debug_assert((TLB_FLAGS_MASK & ((1 << r) - 1)) == 0);
+#endif
+ return r;
+}
+
typedef tcg_target_ulong TCGArg;
/* Define a type and accessor macros for variables. Using pointer types
@@ -790,15 +860,6 @@ do {\
abort();\
} while (0)
-#ifdef CONFIG_DEBUG_TCG
-# define tcg_debug_assert(X) do { assert(X); } while (0)
-#elif QEMU_GNUC_PREREQ(4, 5)
-# define tcg_debug_assert(X) \
- do { if (!(X)) { __builtin_unreachable(); } } while (0)
-#else
-# define tcg_debug_assert(X) do { (void)(X); } while (0)
-#endif
-
void tcg_add_target_add_op_defs(const TCGTargetOpDef *tdefs);
#if UINTPTR_MAX == UINT32_MAX
--
2.7.4
^ permalink raw reply related [flat|nested] 10+ messages in thread* Re: [Qemu-devel] [PULL v2 0/4] tcg-ish updates
2016-07-06 16:18 [Qemu-devel] [PULL v2 0/4] tcg-ish updates Richard Henderson
` (3 preceding siblings ...)
2016-07-06 16:18 ` [Qemu-devel] [PULL v2 4/4] tcg: Improve the alignment check infrastructure Richard Henderson
@ 2016-07-07 9:28 ` Peter Maydell
4 siblings, 0 replies; 10+ messages in thread
From: Peter Maydell @ 2016-07-07 9:28 UTC (permalink / raw)
To: Richard Henderson; +Cc: QEMU Developers
On 6 July 2016 at 17:18, Richard Henderson <rth@twiddle.net> wrote:
> Three tcg patches, and a build change so that we can still
> use --cpu=i686 on a x86_64 host, now that linux-user has
> the syscall assembly thunks.
>
> Change from v1 to v2 is a one-line revert to patch 3, the
> spilling of constants, which affected arm32.
>
> I had thought that I could move the clearing of mem_coherent
> from temp_load (where it just seems odd), to tcg_reg_alloc_movi
> (where it makes more sense). I don't yet know what path was
> missed, but the change left some registers out of sync, and
> thus later containing garbage. Anyway, reverting it is good
> enough for now.
>
>
> r~
>
>
> The following changes since commit 791b7d2340cfafcac9af7864343cf23504d57804:
>
> Merge remote-tracking branch 'remotes/mst/tags/for_upstream' into staging (2016-07-05 16:48:24 +0100)
>
> are available in the git repository at:
>
> git://github.com/rth7680/qemu.git tags/pull-tcg-20160706
>
> for you to fetch changes up to 1f00b27f17518a1bcb4cedca49eaec96a4d560bd:
>
> tcg: Improve the alignment check infrastructure (2016-07-05 20:50:13 -0700)
>
> ----------------------------------------------------------------
> misc updates
>
Applied, thanks.
-- PMM
^ permalink raw reply [flat|nested] 10+ messages in thread