[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Qemu-devel] [PATCH 11/14] ppc: store CR registers in 32 1-bit regis
From: |
Paolo Bonzini |
Subject: |
Re: [Qemu-devel] [PATCH 11/14] ppc: store CR registers in 32 1-bit registers |
Date: |
Fri, 19 Sep 2014 15:53:14 +0200 |
User-agent: |
Mozilla/5.0 (X11; Linux x86_64; rv:31.0) Gecko/20100101 Thunderbird/31.0 |
Il 18/09/2014 22:25, Tom Musta ha scritto:
> This breaks what you did in patch 5, which used LE bit numbering to
> perform shifts.
Yeah, I change "1 << x" to "8 >> x" in this patch for the fcmp
helpers, but not the others.
> And it breaks other code that uses the old LE
> convention.
I'll fix it like this:
git diff target-ppc/fpu_helper.c
diff --git a/target-ppc/fpu_helper.c b/target-ppc/fpu_helper.c
index da93d12..06e4559 100644
--- a/target-ppc/fpu_helper.c
+++ b/target-ppc/fpu_helper.c
@@ -1018,32 +1018,32 @@ uint32_t helper_ftdiv(uint64_t fra, uint64_t frb)
if (unlikely(float64_is_infinity(fra) ||
float64_is_infinity(frb) ||
float64_is_zero(frb))) {
- fe_flag = 1;
- fg_flag = 1;
+ fe_flag = 8 >> CRF_EQ;
+ fg_flag = 8 >> CRF_GT;
} else {
int e_a = ppc_float64_get_unbiased_exp(fra);
int e_b = ppc_float64_get_unbiased_exp(frb);
if (unlikely(float64_is_any_nan(fra) ||
float64_is_any_nan(frb))) {
- fe_flag = 1;
+ fe_flag = 8 >> CRF_EQ;
} else if ((e_b <= -1022) || (e_b >= 1021)) {
- fe_flag = 1;
+ fe_flag = 8 >> CRF_EQ;
} else if (!float64_is_zero(fra) &&
(((e_a - e_b) >= 1023) ||
((e_a - e_b) <= -1021) ||
(e_a <= -970))) {
- fe_flag = 1;
+ fe_flag = 8 >> CRF_EQ;
}
if (unlikely(float64_is_zero_or_denormal(frb))) {
/* XB is not zero because of the above check and */
/* so must be denormalized. */
- fg_flag = 1;
+ fg_flag = 8 >> CRF_GT;
}
}
- return 0x8 | (fg_flag ? 4 : 0) | (fe_flag ? 2 : 0);
+ return (8 >> CRF_LT) | fg_flag | fe_flag;
}
uint32_t helper_ftsqrt(uint64_t frb)
and similarly for ftsqrt.
Paolo
> There are some other places in helper where env->crf[*] was still being set.
> Here are the ones that I found:
>
> diff --git a/target-ppc/fpu_helper.c b/target-ppc/fpu_helper.c
> index 3f656e5..e624f97 100644
> --- a/target-ppc/fpu_helper.c
> +++ b/target-ppc/fpu_helper.c
> @@ -2141,7 +2141,8 @@ void helper_##op(CPUPPCState *env, uint32_t opcode)
> \
> } \
> } \
> \
> - env->crf[BF(opcode)] = 0x8 | (fg_flag ? 4 : 0) | (fe_flag ? 2 : 0); \
> + ppc_set_crf(env, BF(opcode), \
> + 0x8 | (fg_flag ? 4 : 0) | (fe_flag ? 2 : 0)); \
> }
>
> VSX_TDIV(xstdivdp, 1, float64, VsrD(0), -1022, 1023, 52)
> @@ -2195,7 +2196,8 @@ void helper_##op(CPUPPCState *env, uint32_t opcode)
> \
> } \
> } \
> \
> - env->crf[BF(opcode)] = 0x8 | (fg_flag ? 4 : 0) | (fe_flag ? 2 : 0); \
> + ppc_set_crf(env, BF(opcode), \
> + 0x8 | (fg_flag ? 4 : 0) | (fe_flag ? 2 : 0)); \
> }
>
> VSX_TSQRT(xstsqrtdp, 1, float64, VsrD(0), -1022, 52)
> @@ -2358,7 +2360,7 @@ void helper_##op(CPUPPCState *env, uint32_t opcode)
> \
> \
> env->fpscr &= ~(0x0F << FPSCR_FPRF); \
> env->fpscr |= cc << FPSCR_FPRF; \
> - env->crf[BF(opcode)] = cc; \
> + ppc_set_crf(env, BF(opcode), cc); \
> \
> helper_float_check_status(env); \
> }
> @@ -2450,7 +2452,8 @@ void helper_##op(CPUPPCState *env, uint32_t opcode)
> \
> \
> putVSR(xT(opcode), &xt, env); \
> if ((opcode >> (31-21)) & 1) { \
> - env->crf[6] = (all_true ? 0x8 : 0) | (all_false ? 0x2 : 0); \
> + ppc_set_crf(env, 6, \
> + (all_true ? 0x8 : 0) | (all_false ? 0x2 : 0)); \
> } \
> helper_float_check_status(env); \
> }
>
>
>
> Note that I do not have the capability of testing any of the SPE instructions.
>
- Re: [Qemu-devel] [PATCH 08/14] ppc: introduce gen_op_mfcr/gen_op_mtcr, (continued)
- [Qemu-devel] [PATCH 09/14] ppc: introduce ppc_get_crf and ppc_set_crf, Paolo Bonzini, 2014/09/15
- [Qemu-devel] [PATCH 10/14] ppc: use movcond for isel, Paolo Bonzini, 2014/09/15
- [Qemu-devel] [PATCH 11/14] ppc: store CR registers in 32 1-bit registers, Paolo Bonzini, 2014/09/15
- [Qemu-devel] [PATCH 12/14] ppc: use movcond to implement evsel, Paolo Bonzini, 2014/09/15
- [Qemu-devel] [PATCH 13/14] ppc: inline ppc_set_crf when clearer, Paolo Bonzini, 2014/09/15
- [Qemu-devel] [PATCH 14/14] ppc: dump all 32 CR bits, Paolo Bonzini, 2014/09/15
- Re: [Qemu-devel] [PATCH v2 00/14] TCG ppc speedups, Tom Musta, 2014/09/18