* Re: [Qemu-devel] fix evsrwu and others for powerpcspe
@ 2018-02-10 7:25 ` Michael Tokarev
0 siblings, 0 replies; 7+ messages in thread
From: Michael Tokarev @ 2018-02-10 7:25 UTC (permalink / raw)
To: info, qemu-trivial, qemu-devel
26.12.2017 14:29, info@mobile-stream.com wrote:
> GEN_SPEOP_ARITH2 declares non-local temporaries then calls gen_op_evsrwu/evsrws/evslw which generates branch insn.
I'm Cc'ing qemu-devel@ for review, since I don't know much about how ppc works.
Thanks,
/mjt
> --- target/ppc/translate/spe-impl.inc.c.orig
> +++ target/ppc/translate/spe-impl.inc.c
> @@ -158,8 +158,8 @@
> gen_exception(ctx, POWERPC_EXCP_SPEU); \
> return; \
> } \
> - t0 = tcg_temp_new_i32(); \
> - t1 = tcg_temp_new_i32(); \
> + t0 = tcg_temp_local_new_i32(); \
> + t1 = tcg_temp_local_new_i32(); \
> \
> tcg_gen_trunc_tl_i32(t0, cpu_gpr[rA(ctx->opcode)]); \
> tcg_gen_trunc_tl_i32(t1, cpu_gpr[rB(ctx->opcode)]); \
>
>
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [Qemu-trivial] [Qemu-devel] fix evsrwu and others for powerpcspe
2018-02-10 7:25 ` [Qemu-devel] " Michael Tokarev
@ 2018-02-10 7:44 ` Thomas Huth
-1 siblings, 0 replies; 7+ messages in thread
From: Thomas Huth @ 2018-02-10 7:44 UTC (permalink / raw)
To: Michael Tokarev, info, qemu-trivial, qemu-devel
Cc: qemu-ppc, David Gibson, Richard Henderson, Laurent Vivier
On 10.02.2018 08:25, Michael Tokarev wrote:
> 26.12.2017 14:29, info@mobile-stream.com wrote:
>> GEN_SPEOP_ARITH2 declares non-local temporaries then calls gen_op_evsrwu/evsrws/evslw which generates branch insn.
>
> I'm Cc'ing qemu-devel@ for review, since I don't know much about how ppc works.
For PPC related stuff, please also always send a CC: to
qemu-ppc@nongnu.org (or even better, use the scripts/get_maintainer.pl
script to get the CC: list right).
Also, your patch is missing a proper "Signed-off-by" line. Please have a
look at https://wiki.qemu.org/Contribute/SubmitAPatch for details.
Thanks,
Thomas
>
>> --- target/ppc/translate/spe-impl.inc.c.orig
>> +++ target/ppc/translate/spe-impl.inc.c
>> @@ -158,8 +158,8 @@
>> gen_exception(ctx, POWERPC_EXCP_SPEU); \
>> return; \
>> } \
>> - t0 = tcg_temp_new_i32(); \
>> - t1 = tcg_temp_new_i32(); \
>> + t0 = tcg_temp_local_new_i32(); \
>> + t1 = tcg_temp_local_new_i32(); \
>> \
>> tcg_gen_trunc_tl_i32(t0, cpu_gpr[rA(ctx->opcode)]); \
>> tcg_gen_trunc_tl_i32(t1, cpu_gpr[rB(ctx->opcode)]); \
>>
>>
>
>
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [Qemu-devel] fix evsrwu and others for powerpcspe
@ 2018-02-10 7:44 ` Thomas Huth
0 siblings, 0 replies; 7+ messages in thread
From: Thomas Huth @ 2018-02-10 7:44 UTC (permalink / raw)
To: Michael Tokarev, info, qemu-trivial, qemu-devel
Cc: qemu-ppc, David Gibson, Richard Henderson, Laurent Vivier
On 10.02.2018 08:25, Michael Tokarev wrote:
> 26.12.2017 14:29, info@mobile-stream.com wrote:
>> GEN_SPEOP_ARITH2 declares non-local temporaries then calls gen_op_evsrwu/evsrws/evslw which generates branch insn.
>
> I'm Cc'ing qemu-devel@ for review, since I don't know much about how ppc works.
For PPC related stuff, please also always send a CC: to
qemu-ppc@nongnu.org (or even better, use the scripts/get_maintainer.pl
script to get the CC: list right).
Also, your patch is missing a proper "Signed-off-by" line. Please have a
look at https://wiki.qemu.org/Contribute/SubmitAPatch for details.
Thanks,
Thomas
>
>> --- target/ppc/translate/spe-impl.inc.c.orig
>> +++ target/ppc/translate/spe-impl.inc.c
>> @@ -158,8 +158,8 @@
>> gen_exception(ctx, POWERPC_EXCP_SPEU); \
>> return; \
>> } \
>> - t0 = tcg_temp_new_i32(); \
>> - t1 = tcg_temp_new_i32(); \
>> + t0 = tcg_temp_local_new_i32(); \
>> + t1 = tcg_temp_local_new_i32(); \
>> \
>> tcg_gen_trunc_tl_i32(t0, cpu_gpr[rA(ctx->opcode)]); \
>> tcg_gen_trunc_tl_i32(t1, cpu_gpr[rB(ctx->opcode)]); \
>>
>>
>
>
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [Qemu-trivial] [Qemu-devel] fix evsrwu and others for powerpcspe
2018-02-10 7:25 ` [Qemu-devel] " Michael Tokarev
@ 2018-02-10 17:19 ` Richard Henderson
-1 siblings, 0 replies; 7+ messages in thread
From: Richard Henderson @ 2018-02-10 17:19 UTC (permalink / raw)
To: Michael Tokarev, info, qemu-trivial, qemu-devel,
qemu-ppc@nongnu.org
On 02/09/2018 11:25 PM, Michael Tokarev wrote:
> 26.12.2017 14:29, info@mobile-stream.com wrote:
>> GEN_SPEOP_ARITH2 declares non-local temporaries then calls gen_op_evsrwu/evsrws/evslw which generates branch insn.
>
> I'm Cc'ing qemu-devel@ for review, since I don't know much about how ppc works.
>
> Thanks,
>
> /mjt
>
>> --- target/ppc/translate/spe-impl.inc.c.orig
>> +++ target/ppc/translate/spe-impl.inc.c
>> @@ -158,8 +158,8 @@
>> gen_exception(ctx, POWERPC_EXCP_SPEU); \
>> return; \
>> } \
>> - t0 = tcg_temp_new_i32(); \
>> - t1 = tcg_temp_new_i32(); \
>> + t0 = tcg_temp_local_new_i32(); \
>> + t1 = tcg_temp_local_new_i32(); \
While that's one possible solution, it would be better to rewrite the helpers
not to use branches. E.g.
tcg_gen_andi_i32(t0, arg2, 0x1f);
tcg_gen_andi_i32(t1, arg2, 0x20);
tcg_gen_shr_i32(ret, arg1, t0);
tcg_gen_movi_i32(t0, 0);
tcg_gen_movcond_i32(TCG_COND_NE, ret, t1, t0, t0, ret);
r~
^ permalink raw reply [flat|nested] 7+ messages in thread* Re: [Qemu-devel] fix evsrwu and others for powerpcspe
@ 2018-02-10 17:19 ` Richard Henderson
0 siblings, 0 replies; 7+ messages in thread
From: Richard Henderson @ 2018-02-10 17:19 UTC (permalink / raw)
To: Michael Tokarev, info, qemu-trivial, qemu-devel,
qemu-ppc@nongnu.org
On 02/09/2018 11:25 PM, Michael Tokarev wrote:
> 26.12.2017 14:29, info@mobile-stream.com wrote:
>> GEN_SPEOP_ARITH2 declares non-local temporaries then calls gen_op_evsrwu/evsrws/evslw which generates branch insn.
>
> I'm Cc'ing qemu-devel@ for review, since I don't know much about how ppc works.
>
> Thanks,
>
> /mjt
>
>> --- target/ppc/translate/spe-impl.inc.c.orig
>> +++ target/ppc/translate/spe-impl.inc.c
>> @@ -158,8 +158,8 @@
>> gen_exception(ctx, POWERPC_EXCP_SPEU); \
>> return; \
>> } \
>> - t0 = tcg_temp_new_i32(); \
>> - t1 = tcg_temp_new_i32(); \
>> + t0 = tcg_temp_local_new_i32(); \
>> + t1 = tcg_temp_local_new_i32(); \
While that's one possible solution, it would be better to rewrite the helpers
not to use branches. E.g.
tcg_gen_andi_i32(t0, arg2, 0x1f);
tcg_gen_andi_i32(t1, arg2, 0x20);
tcg_gen_shr_i32(ret, arg1, t0);
tcg_gen_movi_i32(t0, 0);
tcg_gen_movcond_i32(TCG_COND_NE, ret, t1, t0, t0, ret);
r~
^ permalink raw reply [flat|nested] 7+ messages in thread