Multiplication of two fractional word elements is not correct when sign
extension/promotion is needed. This change fixes it by adding correct
casts from unsigned to signed values.
In addition, the tests (dpaq_sa_l_w.c and dpsq_sa_l_w.c) have been extended
to trigger the current issue.
Signed-off-by: Petar Jovanovic <petar.jovanovic@imgtec.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
This change corrects rounding and saturation of Q31 fractional value in
mipsdsp_trunc16_sat16_round(). Overflow detection was incorrect for the
corner case for PRECRQ_RS.PH, and this test case is also part of the change.
Signed-off-by: Petar Jovanovic <petar.jovanovic@imgtec.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
Multiplication of Q15 fractional halfword vectors was incorrect in the
previous implementation of mipsdsp_rndq15_mul_q15_q15. It failed to take
element signs into account. This change fixes it, and it adds a test case
for it.
The change also removes unnecessary cast in the function
mipsdsp_mul_q15_q15_overflowflag21().
Signed-off-by: Petar Jovanovic <petar.jovanovic@imgtec.com>
Reviewed-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
This is a small follow-up change to "fix incorrect behaviour for INSV".
It includes two minor modifications:
- sizefilter is constant so it can be moved inside of the block,
- several lines of the code are replaced with a call to deposit64.
No functional change.
Signed-off-by: Petar Jovanovic <petar.jovanovic@imgtec.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
First we need to clear the bit and then we set the given value.
Instruction ADDSC sets the bit and instruction ADDWC uses this bit.
Signed-off-by: Petar Jovanovic <petar.jovanovic@imgtec.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
This change makes sure that modifications of pos field in the DSPControl
register do not trash other bits in the register. This bug can be triggered
with the additional test case in mips32-dsp/extpdp.c in this commit.
In addition to this, this change corrects incorrect calculation of the mask
for EXTPDP.
Signed-off-by: Petar Jovanovic <petar.jovanovic@imgtec.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
The mask for EXTP instruction when size=31 has not been correctly
calculated.
The test (mips32-dsp/extp.c) has been extended to include the case that
triggers the issue.
Signed-off-by: Petar Jovanovic <petar.jovanovic@imgtec.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
Corner case for INSV instruction when size=32 has not been correctly
implemented. The mask for size should be one bit wider, and preparing the
filter variable should be aware of this case too.
The test for INSV has been extended to include the case that triggers the
bug.
Signed-off-by: Petar Jovanovic <petar.jovanovic@imgtec.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
This change corrects and simplifies how discard is calculated for shift
left logical vector instructions. It is used to detect overflow and set bit
22 in the DSPControl register.
The existing tests (shll_ph.c, shll_qb.c) are extended with the corner cases
that expose incorrectness in the previous implementation.
Signed-off-by: Petar Jovanovic <petar.jovanovic@imgtec.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
The operands for MAQ_SA_W.PHL/MAQ_SA_W.PHR must in specified format.
Otherwise, the results are unpredictable. Once the operands were corrected
in the tests (part of this change), a bug in mipsdsp_mul_q15_q15 became
visible.
This change corrects the tests for MAQ_SA_W.PHL/MAQ_SA_W.PHR and fixes
sign-related issue in mipsdsp_mul_q15_q15. It also removes unnecessary
comment.
Signed-off-by: Petar Jovanovic <petar.jovanovic@imgtec.com>
Fix for rndrashift_short_acc to set correct value to higher 64 bits.
This change also corrects conditions when bit 23 of the DSPControl register
is set.
The existing test files have been extended with several examples that
trigger the issues. One bug/example in the test file for EXTR_RS_W has been
found and reported by Klaus Peichl.
Signed-off-by: Petar Jovanovic <petar.jovanovic@imgtec.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
The previous implementation incorrectly used same macro to detect overflow
for addition and subtraction. This patch makes distinction between these
two, and creates separate macros. The affected routines are changed
accordingly.
This change also includes additions to the existing tests for SUBQ_S_PH and
SUBQ_S_W that would trigger the fixed issue, and it removes dead code from
the test file. The last test case in subq_s_w.c is a bug found/reported/
isolated by Klaus Peichl from Dolby.
Signed-off-by: Petar Jovanovic <petar.jovanovic@imgtec.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
Correct sign-propagation before multiplication in MULQ_W helper.
The change also fixes previously incorrect expected values in the
tests for MULQ_RS.W and MULQ_S.W.
Signed-off-by: Petar Jovanovic <petarj@mips.com>
Richard Henderson <rth@twiddle.net>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
The change corrects sign-related issue with MULQ_S.PH. It also includes
extension to the already existing test which will trigger the issue.
Signed-off-by: Petar Jovanovic <petarj@mips.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
DSP instruction from the (d)append sub-class can be implemented with
TCG. Use a different function for these instructions are they are quite
different from compare-pick sub-class.
Fix BALIGN instruction for negative value, where the value should be
zero-extended before being shift to the right.
Reviewed-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
Instead of playing with bit shifting, add two unions (one for 32-bit
values, one for 64-bit ones) to access all the DSP elements with the
correct type.
This make the code easier to read and less error prone, and allow GCC
to vectorize the code in some cases.
Reviewed-by: Eric Johnson <ericj@mips.com>
Reviewed-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
Helper function for dpa_w_ph, dpax_w_ph, dps_w_ph and dpsx_w_ph incorrectly
defines halfword vector elements as unsigned values. This results in wrong
output which is not triggered in the tests as they also follow this logic.
Signed-off-by: Petar Jovanovic <petarj@mips.com>
Reviewed-by: Eric Johnson <ericj@mips.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
The change removes some unnecessary and incorrect code for EXTR_S.H.
Further, it corrects the mask for shift value in the EXTR_ instructions. It also
extends the existing tests so they trigger the issues corrected with the change.
Signed-off-by: Petar Jovanovic <petarj@mips.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
Upper 4 bits of ccond (bits 31..28 ) of DSPControl register are not used in
the MIPS32 architecture. They are used in the MIPS64 architecture. For MIPS32
these bits must be written as zero, and return zero on read.
The change fixes writes (WRDSP) and reads (RDDSP) to the register. It also fixes
the tests that use these instructions, and makes them smaller and simpler.
Signed-off-by: Petar Jovanovic <petarj@mips.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
helper_shilo has not been shifting an accumulator value correctly for negative
values in 'shift' field. Minor optimization for shift=0 case.
This change also adds tests that will trigger issue and check for regressions.
Signed-off-by: Petar Jovanovic <petarj@mips.com>
Reviewed-by: Richard Henderson <rth@twiddle.net>
Reviewed-by: Eric Johnson <ericj@mips.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
Content of register rs should be shifted for pos before applying a mask.
This change contains both fix for the instruction and to the existing test.
Signed-off-by: Petar Jovanovic <petarj@mips.com>
Reviewed-by: Eric Johnson <ericj@mips.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
Fix build on a 32 bit host:
CC mips-softmmu/target-mips/dsp_helper.o
/src/qemu/target-mips/dsp_helper.c: In function 'helper_dextr_rs_w':
/src/qemu/target-mips/dsp_helper.c:3556: error: integer constant is too large for 'long' type
/src/qemu/target-mips/dsp_helper.c: In function 'helper_extr_s_h':
/src/qemu/target-mips/dsp_helper.c:3656: error: integer constant is too large for 'long' type
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
Add internal functions using by MIPS ASE DSP instructions.
Signed-off-by: Jia Liu <proljc@gmail.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
