From: Richard Henderson <richard.henderson@linaro.org>
To: "Bruno Larsen (billionai)" <bruno.larsen@eldorado.org.br>,
qemu-devel@nongnu.org
Cc: farosas@linux.ibm.com, luis.pires@eldorado.org.br,
lucas.araujo@eldorado.org.br, fernando.valle@eldorado.org.br,
qemu-ppc@nongnu.org, matheus.ferst@eldorado.org.br,
david@gibson.dropbear.id.au
Subject: Re: [PATCH 07/11] target/ppc: added KVM fallback to fpscr manipulation
Date: Wed, 12 May 2021 13:20:14 -0500 [thread overview]
Message-ID: <a9ac8b35-faae-167d-ff78-4b7ddd8f75c9@linaro.org> (raw)
In-Reply-To: <20210512140813.112884-8-bruno.larsen@eldorado.org.br>
On 5/12/21 9:08 AM, Bruno Larsen (billionai) wrote:
> diff --git a/target/ppc/kvm.c b/target/ppc/kvm.c
> index 104a308abb..a8a720eb48 100644
> --- a/target/ppc/kvm.c
> +++ b/target/ppc/kvm.c
> @@ -2947,3 +2947,17 @@ bool kvm_arch_cpu_check_are_resettable(void)
> {
> return true;
> }
> +
> +void kvmppc_store_fpscr(CPUPPCState *env, uint64_t arg, uint32_t mask)
> +{
> + CPUState *cs = env_cpu(env);
> + struct kvm_one_reg reg;
> + int ret;
> + reg.id = KVM_REG_PPC_FPSCR;
> + reg.addr = (uintptr_t) &env->fpscr;
> + ret = kvm_vcpu_ioctl(cs, KVM_SET_ONE_REG, ®);
> + if (ret < 0)
> + {
> + fprintf(stderr, "Unable to set FPSCR to KVM: %s", strerror(errno));
> + }
> +}
This is all unnecessary. All you need to do is store to env->fpscr and the
value will be synced back with kvm_put_fp.
I'll note that some of the trouble you may be having with extracting
helper_store_fpscr to a ppc_store_fpscr function is due to an existing bug in
the tcg code:
Storing to fpscr should *never* raise an exception -- see MTFSF, MTFSB0,
MTFSB1. Thus the mucking about with cs->exception_index and env->error_code is
incorrect.
In addition, the masking is being done weirdly and could use a complete overhaul.
This could all be rewritten as
-- %< -- cpu.h
/* Invalid operation exception summary */
- #define fpscr_ix ((env->fpscr) & ((1 << FPSCR_VXSNAN) ...
+ #define FPSCR_IX ((1 << FPSCR_VXSNAN) | ...)
-- %< -- cpu.c
// move fpscr_set_rounding_mode here
void ppc_store_fpscr(CPUPPCState *env, target_ulong val)
{
/* Recompute exception summary state. */
val &= ~(FP_VX | FP_FEX);
if (val & FPSCR_IX) {
val |= FP_VX;
}
if ((val >> FPSCR_XX) & (val >> FPSCR_XE) & 0x1f) {
val |= FP_FEX;
}
env->fpscr = val;
if (tcg_enabled()) {
fpscr_set_rounding_mode(env);
}
}
-- %< -- fpu_helper.c
void helper_store_fpscr(CPUPPCState *env, target_ulong val,
uint32_t nibbles)
{
target_ulong mask = 0;
/* TODO: Push this expansion back to translation time. */
for (int i = 0; i < sizeof(target_ulong) * 2; ++i) {
if (nibbles & (1 << i)) {
mask |= (target_ulong)0xf << (4 * i);
}
}
val = (val & mask) | (env->fpscr & ~mask);
ppc_store_fpscr(env, val);
}
void helper_fpscr_clrbit(CPUPPCState *env, uint32_t bit)
{
uint32_t mask = 1u << bit;
if (env->fpscr & mask) {
ppc_store_fpscr(env, env->fpscr & ~mask);
}
}
void helper_fpscr_setbit(CPUPPCState *env, uint32_t bit)
{
uint32_t mask = 1u << bit;
if (!(env->fpscr & mask)) {
ppc_store_fpscr(env, env->fpscr | mask);
}
}
There are a couple of other uses of fpscr_set_rounding_mode, where the
softfloat value is changed temporarily (do_fri, VSX_ROUND). These should
simply save the previous softfloat value (using get_float_rounding_mode) around
the operation instead of re-computing from fpscr.
Which leaves us with exactly one use of fpscr_set_rounding_mode, which can then
be moved to cpu.c next to ppc_store_fpscr.
r~
next prev parent reply other threads:[~2021-05-12 18:23 UTC|newest]
Thread overview: 59+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-05-12 14:08 [RFC PATCH 00/11] target/ppc: add support to disable-tcg Bruno Larsen (billionai)
2021-05-12 14:08 ` [PATCH 01/11] target/ppc: created ppc_{store, get}_vscr for generic vscr usage Bruno Larsen (billionai)
2021-05-12 16:48 ` [PATCH 01/11] target/ppc: created ppc_{store,get}_vscr " Richard Henderson
2021-05-13 3:48 ` David Gibson
2021-05-12 14:08 ` [PATCH 02/11] target/ppc: moved ppc_store_sdr1 to cpu.c Bruno Larsen (billionai)
2021-05-12 16:54 ` Richard Henderson
2021-05-13 3:50 ` David Gibson
2021-05-31 18:17 ` Bruno Piazera Larsen
2021-06-07 3:50 ` David Gibson
2021-05-12 14:08 ` [PATCH 03/11] target/ppc: moved ppc_cpu_dump_state to cpu_init.c Bruno Larsen (billionai)
2021-05-12 16:58 ` Richard Henderson
2021-05-12 17:00 ` Richard Henderson
2021-05-13 3:51 ` David Gibson
2021-05-12 14:08 ` [PATCH 04/11] target/ppc: moved ppc_store_msr into gdbstub.c Bruno Larsen (billionai)
2021-05-12 17:05 ` Richard Henderson
2021-05-13 3:52 ` David Gibson
2021-05-12 14:08 ` [PATCH 05/11] target/ppc: moved ppc_store_lpcr to cpu.c Bruno Larsen (billionai)
2021-05-12 17:07 ` Richard Henderson
2021-05-13 3:53 ` David Gibson
2021-05-12 14:08 ` [PATCH 06/11] target/ppc: updated vscr manipulation in machine.c Bruno Larsen (billionai)
2021-05-12 17:08 ` Richard Henderson
2021-05-13 3:54 ` David Gibson
2021-05-12 14:08 ` [PATCH 07/11] target/ppc: added KVM fallback to fpscr manipulation Bruno Larsen (billionai)
2021-05-12 18:20 ` Richard Henderson [this message]
2021-05-12 19:15 ` Bruno Piazera Larsen
2021-05-13 16:36 ` Bruno Piazera Larsen
2021-05-13 22:45 ` Richard Henderson
2021-05-14 11:12 ` Bruno Piazera Larsen
2021-05-12 14:08 ` [RFC PATCH 08/11] target/ppc: wrapped some TCG only logic with ifdefs Bruno Larsen (billionai)
2021-05-12 18:33 ` Richard Henderson
2021-05-12 18:57 ` Bruno Piazera Larsen
2021-05-14 13:29 ` Bruno Piazera Larsen
2021-05-14 14:44 ` Richard Henderson
2021-05-14 16:22 ` Bruno Piazera Larsen
2021-05-17 4:10 ` David Gibson
2021-05-17 16:07 ` Richard Henderson
2021-05-17 4:00 ` David Gibson
2021-05-24 18:01 ` Bruno Piazera Larsen
2021-05-24 20:03 ` Richard Henderson
2021-05-12 14:08 ` [PATCH 09/11] include/exec: added functions to the stubs in exec-all.h Bruno Larsen (billionai)
2021-05-12 18:34 ` Richard Henderson
2021-05-13 14:03 ` Lucas Mateus Martins Araujo e Castro
2021-05-13 23:44 ` Richard Henderson
2021-05-17 3:58 ` David Gibson
2021-05-17 16:59 ` Lucas Mateus Martins Araujo e Castro
2021-05-17 18:02 ` Richard Henderson
2021-05-18 18:45 ` Bruno Piazera Larsen
2021-05-24 3:18 ` David Gibson
2021-05-17 3:55 ` David Gibson
2021-05-17 11:07 ` Bruno Piazera Larsen
2021-05-24 3:15 ` David Gibson
2021-05-12 14:08 ` [RFC PATCH 10/11] target/ppc: created tcg-stub.c file Bruno Larsen (billionai)
2021-05-12 18:39 ` Richard Henderson
2021-05-12 19:09 ` Bruno Piazera Larsen
2021-05-14 18:07 ` Bruno Piazera Larsen
2021-05-13 3:59 ` David Gibson
2021-05-13 12:56 ` Bruno Piazera Larsen
2021-05-17 3:53 ` David Gibson
2021-05-12 14:08 ` [PATCH 11/11] target/ppc: updated meson.build to support disable-tcg Bruno Larsen (billionai)
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=a9ac8b35-faae-167d-ff78-4b7ddd8f75c9@linaro.org \
--to=richard.henderson@linaro.org \
--cc=bruno.larsen@eldorado.org.br \
--cc=david@gibson.dropbear.id.au \
--cc=farosas@linux.ibm.com \
--cc=fernando.valle@eldorado.org.br \
--cc=lucas.araujo@eldorado.org.br \
--cc=luis.pires@eldorado.org.br \
--cc=matheus.ferst@eldorado.org.br \
--cc=qemu-devel@nongnu.org \
--cc=qemu-ppc@nongnu.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).