qemu-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Qemu-devel] [PATCH] softfloat: make USE_SOFTFLOAT_STRUCT_TYPES comp


From: Peter Maydell
Subject: Re: [Qemu-devel] [PATCH] softfloat: make USE_SOFTFLOAT_STRUCT_TYPES compile
Date: Tue, 20 Mar 2012 16:11:00 +0000

On 20 March 2012 15:41, Andreas Färber <address@hidden> wrote:
> Tested the default, non-struct version only.
>
> Juan, would it make sense to add a configure option for this #define,
> similar to --enable-debug-tcg?

We'd need to fix some of the other targets:
target-alpha/machine.c:25: error: invalid operands to binary - (have
‘uint64_t (*)[31]’ and ‘struct float64 (*)[31]’)
target-m68k/helper.c:540: error: conflicting types for ‘helper_f64_to_i32’
target-microblaze/op_helper.c:447: error: incompatible types when
assigning to type ‘uint32_t’ from type ‘float32’
target-mips/op_helper.c:2487: error: incompatible type for argument 1
of ‘float64_sqrt’
target-ppc/op_helper.c:1236: error: incompatible types when assigning
to type ‘uint64_t’ from type ‘float64’
gdbstub.c:1177: error: incompatible type for argument 2 of ‘stl_le_p’
target-s390x/op_helper.c:1056: error: incompatible types when
assigning to type ‘uint32_t’ from type ‘float32’
target-unicore32/translate.c:2094: error: aggregate value used where
an integer was expected

(one sample error message only from each; eg mips has tons)

A quick look at gcc object files suggests that we do generate
slightly worse code sometimes (a few extra insns here and there,
esp. for float16). I think we could reasonably turn on this
debug by default if --enable-debug is turned on (the way that
--enable-debug-tcg is turned on by --enable-debug).

I think we should either have this type checking happen when
you use --enable-debug or we should say we flat out don't care
about float32/uint32_t mismatches; the current situation just
means we accumulate errors without noticing.

-- PMM



reply via email to

[Prev in Thread] Current Thread [Next in Thread]