ISA v3.1 changed some VSX instructions behavior by changing what the
other words/doubleword in the result should contain when the result is
only one word/doubleword. e.g. xsmaxdp operates on doubleword 0 and
saves the result also in doubleword 0.
Before, the second doubleword result was undefined according to the
ISA, but now it's stated that it should be zeroed.
Even tough the result was undefined before, hardware implementing these
instructions already filled these fields with 0s. Changing every ISA
version in QEMU to this behavior makes the results match what happens
in hardware.
Signed-off-by: Víctor Colombo <victor.colombo@eldorado.org.br>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20220204181944.65063-1-victor.colombo@eldorado.org.br>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
The non-signalling versions of VSX scalar convert to shorter/longer
precision insns doesn't silence SNaNs in the hardware. To better match
this behavior, use the non-arithmatic conversion of helper_todouble
instead of float32_to_float64. A test is added to prevent future
regressions.
Signed-off-by: Matheus Ferst <matheus.ferst@eldorado.org.br>
Message-Id: <20211228120310.1957990-1-matheus.ferst@eldorado.org.br>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
PPC instruction xsmaxcdp, xsmincdp, xsmaxjdp, and xsminjdp are using
vector registers when they should be using VSX ones. This happens
because the instructions are using GEN_VSX_HELPER_R3, which adds 32
to the register numbers, effectively making them vector registers.
This patch fixes it by changing these instructions to use
GEN_VSX_HELPER_X3.
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Victor Colombo <victor.colombo@eldorado.org.br>
Message-Id: <20211213120958.24443-2-victor.colombo@eldorado.org.br>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
When computing the predicate "is this value currently formatted
for single precision", we do not want to round the value according
to the current rounding mode, nor perform a floating-point equality.
We want to see if the N bits that make up single-precision are the
only ones set within the register, and then a bitwise equality.
Fixes a bug in which a single-precision NaN is considered !SP,
because float64_eq(nan, nan) is always false.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20211119160502.17432-35-richard.henderson@linaro.org>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
There is no double-rounding bug here, because the result is
merely an estimate to within 1 part in 256, but perform the
operation with float64r32_div for consistency.
Use float_flag_invalid_snan instead of recomputing the
snan-ness of the operand.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20211119160502.17432-34-richard.henderson@linaro.org>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
There is no double-rounding bug here, because the result is
merely an estimate to within 1 part in 32, but perform the
operation with float64r32_div for consistency.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20211119160502.17432-33-richard.henderson@linaro.org>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
Use float64r32_mul. Fixes a double-rounding issue with performing
the compuation in float64 and then rounding afterward.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20211119160502.17432-32-richard.henderson@linaro.org>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
Use float64r32_{add,sub,div}. Fixes a double-rounding issue with
performing the compuation in float64 and then rounding afterward.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20211119160502.17432-31-richard.henderson@linaro.org>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
Use float64r32_sqrt. Fixes a double-rounding issue with performing
the compuation in float64 and then rounding afterward.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20211119160502.17432-30-richard.henderson@linaro.org>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
Use float64r32_muladd. Fixes a double-rounding issue with performing
the compuation in float64 and then rounding afterward.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20211119160502.17432-29-richard.henderson@linaro.org>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
Use float_flag_invalid_snan instead of recomputing
the snan-ness of the operand.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20211119160502.17432-27-richard.henderson@linaro.org>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
Use float_flag_invalid_snan instead of recomputing
the snan-ness of the operand.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20211119160502.17432-26-richard.henderson@linaro.org>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
Now that vxsqrt and vxsnan are computed directly by softfloat,
we don't need to recompute it. Split out float_invalid_op_sqrt
to be used in several places. This fixes VSX_SQRT, which did
not order its tests correctly to eliminate NaN with sign set.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20211119160502.17432-25-richard.henderson@linaro.org>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
We only needed one ieee arithmetic operation to raise
exceptions. To convert back to register form, we can
use our simpler non-arithmetic function.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20211119160502.17432-24-richard.henderson@linaro.org>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
Now that vxsnan is computed directly by softfloat,
we don't need to recompute it.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20211119160502.17432-23-richard.henderson@linaro.org>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
Calling helper_frsp directly from other helpers generates
the incorrect retaddr. Split out a helper that takes the
retaddr as a parameter.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20211119160502.17432-22-richard.henderson@linaro.org>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
We will process flags other than in valid in helper_float_check_status,
which is invoked after the writeback to FRT.
Fixes a bug in which FRT is not written when OE/UE/XE are enabled.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20211119160502.17432-21-richard.henderson@linaro.org>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
Create a common function for all of the madd helpers.
Let the compiler tail call or inline as it chooses.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20211119160502.17432-20-richard.henderson@linaro.org>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
Now that vximz, vxisi, and vxsnan are computed directly by
softfloat, we don't need to recompute it. This replaces the
separate float{32,64}_maddsub_update_excp functions with a
single float_invalid_op_madd function.
Fix VSX_MADD by passing sfprf to float_invalid_op_madd,
whereas the previous *_maddsub_update_excp assumed it true.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20211119160502.17432-19-richard.henderson@linaro.org>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
Let float64_round_to_int detect and silence snans.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20211119160502.17432-18-richard.henderson@linaro.org>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
In GEN_FLOAT_B, we called helper_reset_fpstatus immediately
before calling helper_fri*. Therefore get_float_exception_flags
is known to be zero, and this code can be simplified.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20211119160502.17432-17-richard.henderson@linaro.org>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
This is the proper type for the enumeration.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-Id: <20211119160502.17432-16-richard.henderson@linaro.org>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
There's no reason the callers can't tail call to one function.
Leave it up to the compiler either way.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-Id: <20211119160502.17432-15-richard.henderson@linaro.org>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
We were returning nanval for any instance of invalid being set,
but that is an incorrect for VXCVI. This failure can be seen
in the float_convs tests.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20211119160502.17432-14-richard.henderson@linaro.org>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
Now that vxsnan is computed directly by softfloat,
we don't need to recompute it via classes.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20211119160502.17432-13-richard.henderson@linaro.org>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
Fixes a bug in which e.g XE enabled causes inexact to be raised
before the writeback to the architectural register.
All of the users of GEN_FLOAT_B either set set_fprf, or are one
of the convert-to-integer instructions that require this behaviour.
Split out the two gen_helper_* calls in gen_compute_fprf_float64
and protect only the first with set_fprf.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20211119160502.17432-12-richard.henderson@linaro.org>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
Now that vxidi, vxzdz, and vxsnan are computed directly by
softfloat, we don't need to recompute it via classes.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20211119160502.17432-11-richard.henderson@linaro.org>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
Now that vximz and vxsnan are computed directly by
softfloat, we don't need to recompute it via classes.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20211119160502.17432-10-richard.henderson@linaro.org>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
Now that vxisi and vxsnan are computed directly by
softfloat, we don't need to recompute it via classes.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20211119160502.17432-9-richard.henderson@linaro.org>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
mtfsf, mtfsfi and mtfsb1 instructions call helper_float_check_status
after updating the value of FPSCR, but helper_float_check_status
checks fp_status and fp_status isn't updated based on FPSCR and
since the value of fp_status is reset earlier in the instruction,
it's always 0.
Because of this helper_float_check_status would change the FI bit to 0
as this bit checks if the last operation was inexact and
float_flag_inexact is always 0.
These instructions also don't throw exceptions correctly since
helper_float_check_status throw exceptions based on fp_status.
This commit created a new helper, helper_fpscr_check_status that checks
FPSCR value instead of fp_status and checks for a larger variety of
exceptions than do_float_check_status.
Since fp_status isn't used, gen_reset_fpstatus() was removed.
The hardware used to compare QEMU's behavior to was a Power9.
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Lucas Mateus Castro (alqotel) <lucas.araujo@eldorado.org.br>
Message-Id: <20211201163808.440385-2-lucas.araujo@eldorado.org.br>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
Followed the suggested overhaul to store_fpscr logic, and moved it to
cpu.c where it can be accessed in !TCG builds.
The overhaul was suggested because storing a value to fpscr should
never raise an exception, so we could remove all the mess that happened
with POWERPC_EXCP_FP.
We also moved fpscr_set_rounding_mode into cpu.c as it could now be moved
there, and it is needed when a value for the fpscr is being stored
directly.
Suggested-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Bruno Larsen (billionai) <bruno.larsen@eldorado.org.br>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20210527163522.23019-1-bruno.larsen@eldorado.org.br>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
It is preferable to store the current rounding mode and retore from that
than recalculating from fpscr, so we changed the behavior of do_fri and
VSX_ROUND to do it like that.
Suggested-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Bruno Larsen (billionai) <bruno.larsen@eldorado.org.br>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20210521201759.85475-4-bruno.larsen@eldorado.org.br>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
The PowerISA reference states that the comparison operators update the
FPCC, CR and FPSCR and, if VE=1, jump to the exception handler.
Moving the exception-triggering code after the CC update sequence solves
the problem.
Signed-off-by: Giuseppe Musacchio <thatlemon@gmail.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20201112230130.65262-5-thatlemon@gmail.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Since we always perform a comparison between the two operands avoid
checking for NaN unless the result states they're unordered.
Suggested-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Giuseppe Musacchio <thatlemon@gmail.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20201112230130.65262-4-thatlemon@gmail.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Suggested-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Giuseppe Musacchio <thatlemon@gmail.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20201112230130.65262-3-thatlemon@gmail.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
According to the PowerISA v3.1 reference, Table 68 "Actions for xscmpudp
- Part 1: Compare Unordered", whenever one of the two operands is a NaN
the SO bit is set while the other three bits are cleared.
Apply the same change to xscmpuqp.
The respective ordered counterparts are unaffected.
Signed-off-by: Giuseppe Musacchio <thatlemon@gmail.com>
Message-Id: <20201112230130.65262-2-thatlemon@gmail.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
There is no "version 2" of the "Lesser" General Public License.
It is either "GPL version 2.0" or "Lesser GPL version 2.1".
This patch replaces all occurrences of "Lesser GPL version 2" with
"Lesser GPL version 2.1" in comment section.
Signed-off-by: Chetan Pant <chetan4windows@gmail.com>
Message-Id: <20201019061126.3102-1-chetan4windows@gmail.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>
I found that there are many spelling errors in the comments of qemu/target/ppc.
I used spellcheck to check the spelling errors and found some errors in the folder.
Signed-off-by: zhaolichang <zhaolichang@huawei.com>
Reviewed-by: David Edmondson <david.edmondson@oracle.com>
Message-Id: <20201009064449.2336-3-zhaolichang@huawei.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Reviewed-by: Greg Kurz <groug@kaod.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
"Deferred" was misspelled as "differed" in some comments, correct this
typo,
Signed-off-by: BALATON Zoltan <balaton@eik.bme.hu>
Message-Id: <20200214155748.0896B745953@zero.eik.bme.hu>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
There are FPSCR-related defines in target/ppc/cpu.h which can be used in
place of constants and explicit shifts which arguably improve the code a
bit in places.
Signed-off-by: Paul A. Clarke <pc@us.ibm.com>
Message-Id: <1568817169-1721-1-git-send-email-pc@us.ibm.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
The underflow and inexact exceptions are not mutually exclusive.
Check for both of them. Tidy the reset of FPSCR[FI].
Fixes: https://bugs.launchpad.net/bugs/1841442
Reported-by: Paul Clarke <pc@us.ibm.com>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Tested-by: Paul Clarke <pc@us.ibm.com>
Message-Id: <20190826165434.18403-2-richard.henderson@linaro.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
The xscvdpspn instruction implements a non-arithmetic conversion.
In particular, NaNs are not silenced and rounding is not performed.
Rewrite to match the pseudocode for ConvertDPtoSP_NS() in the
Power 3.0B manual.
Signed-off-by: Paul A. Clarke <pc@us.ibm.com>
Message-Id: <1566321964-1447-1-git-send-email-pc@us.ibm.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
[dwg: Replaced description with clearer version from rth]
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
helper_todouble() was not properly converting any denormalized 32 bit
float to 64 bit double.
Fix-suggested-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paul A. Clarke <pc@us.ibm.com>
v2:
- Splitting patch "ppc: Three floating point fixes"; this is just one part.
- Original suggested "fix" was likely flawed. v2 is rewritten by
Richard Henderson (Thanks, Richard!); I reformatted the comments in a
couple of places, compiled, and tested.
Message-Id: <1566250936-14538-1-git-send-email-pc@us.ibm.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
helper_todouble() was not properly converting INFINITY from 32 bit
float to 64 bit double.
(Normalized operand conversion is unchanged, other than indentation.)
Signed-off-by: Paul A. Clarke <pc@us.ibm.com>
Message-Id: <1566242388-9244-1-git-send-email-pc@us.ibm.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
The POWER8 and POWER9 User's Manuals specify the implementation
behavior for what the ISA leaves "undefined" behavior for the
xscvdpspn and xscvdpsp instructions. This patch corrects the QEMU
implementation to match the hardware implementation for that case.
ISA 3.0B has xscvdpspn leaving its result in word 0 of the target register,
with the other words of the target register left "undefined".
The User's Manuals specify:
VSX scalar convert from double-precision to single-precision (xscvdpsp,
xscvdpspn).
VSR[32:63] is set to VSR[0:31].
So, words 0 and 1 both contain the result.
Note: this is important because GCC as of version 8 or so, assumes and takes
advantage of this behavior to optimize the following sequence:
xscvdpspn vs0,vs1
mffprwz r8,f0
ISA 3.0B has xscvdpspn leaving its result in word 0 of the target register,
and mffprwz expecting its input to come from word 1 of the source register.
This sequence fails with QEMU, as a shift is required between those two
instructions. However, since the hardware splats the result to both words 0
and 1 of its output register, the shift is not necessary.
Expect a future revision of the ISA to specify this behavior.
Signed-off-by: Paul A. Clarke <pc@us.ibm.com>
v2
- Splitting patch "ppc: Three floating point fixes"; this is just one part.
- Updated commit message to clarify behavior is documented in User's Manuals.
- Updated commit message to correct which words are in output and source of
xscvdpspn and mffprz.
- No source changes to this part of the original patch.
Message-Id: <1566236601-22954-1-git-send-email-pc@us.ibm.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
ISA 3.0B added a set of Floating-Point Status and Control Register (FPSCR)
instructions: mffsce, mffscdrn, mffscdrni, mffscrn, mffscrni, mffsl.
This patch adds support for 'mffsl'.
'mffsl' is identical to 'mffs', except it only returns mode, status, and enable
bits from the FPSCR.
On CPUs without support for 'mffsl' (below ISA 3.0), the 'mffsl' instruction
will execute identically to 'mffs'.
Note: I renamed FPSCR_RN to FPSCR_RN0 so I could create an FPSCR_RN mask which
is both bits of the FPSCR rounding mode, as defined in the ISA.
I also fixed a typo in the definition of FPSCR_FR.
Signed-off-by: Paul A. Clarke <pc@us.ibm.com>
v4:
- nit: added some braces to resolve a checkpatch complaint.
v3:
- Changed tcg_gen_and_i64 to tcg_gen_andi_i64, eliminating the need for a
temporary, per review from Richard Henderson.
v2:
- I found that I copied too much of the 'mffs' implementation.
The 'Rc' condition code bits are not needed for 'mffsl'. Removed.
- I now free the (renamed) 'tmask' temporary.
- I now bail early for older ISA to the original 'mffs' implementation.
Message-Id: <1565982203-11048-1-git-send-email-pc@us.ibm.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
