Hello all,
Apologies if this has already been addressed, as I am new to this
mailing list and have not had time to search the archives carefully.
I have found and fixed the underlying cause of the illegal instruction
(SIGILL) signals sent to processes that execute some Loongson-specific
floating-point operations (e.g. an madd/msub/nmadd/nmsub instruction
with a quiet NaN or denormalized floating-point input value).
MIPS and Loongson processors only implement the common cases of
floating-point arithmetic operations in hardware. In unusual cases, the
FPU signals an "Unimplemented instruction exception" and relies upon the
OS kernel to emulate the instruction in software.
Linux includes a MIPS floating-point emulator to handle these cases, but
its emulator does not include support for the Loongson variants of
madd/msub/nmadd/nmsub. It also does not include support for the
Loongson operations on "paired-single" floating-point format (vectors of
two single-precision values).
The following patches add support for these instructions to
arch/mips/math-emu. I fixed some other deficiencies in that code as
well, and those fixes are included in separate patches before the
Loongson2f patch.
Note that during testing, I found a mistake in the documentation for the
Loongson 2F. Page 95 of Loongson2FUserGuide.pdf claims that the
paired-single floating-point instructions should use a format (fmt)
field of 11. However, the processor itself does not seem to recognize
instructions with (fmt=11). It seems to be looking for (fmt=22).
That's the case for the Loongson2F in my Yeeloong anyway (which
/proc/cpuinfo identifies as "ICT Loongson-2 V0.3 FPU V0.1").
I also found that the GNU assembler generates instructions with (fmt=22)
for most (but not all) of the paired-single floating-point instructions.
The only exceptions I've found are the Loongson-specific madd.ps,
msub.ps, nmadd.ps, and nmsub.ps instructions, for which GNU as generates
(fmt=11), although that does not work on my Loongson2F. Therefore, I
have also included a patch to GNU binutils (which will also apply to
GDB) to use fmt=22 for those instructions on Loongson2F.
I should note that the documentation for Loongson 2E (not 2F) also
claims that the format field for paired-single instructions should use
fmt=11. I don't know whether this is a mistake or not, since I do not
have access to a Loongson 2E processor with which to test.
Finally, two apologies are in order:
* Ideally I should have added support for the entire set of MIPS 5
floating-point operations, but I did not have time to do that job, nor
do I have a MIPS 5 processor with which to test. Therefore, these
patches address only the Loongson 2F.
* I must admit that I have not tested these patches against the
linux-loongson-community kernel. However, I have tested them against
the linux--2.6.39.1-libre-lemote kernel maintained by Alexandre Oliva,
and there is no significant difference in arch/mips/math-emu between
those two trees.
Comments and suggestions solicited.
Mark
--
You received this message because you are subscribed to the Google Groups
"loongson-dev" group.
To post to this group, send email to loongson-dev@googlegroups.com.
To unsubscribe from this group, send email to
loongson-dev+unsubscr...@googlegroups.com.
For more options, visit this group at
http://groups.google.com/group/loongson-dev?hl=en.
>From c0b01f6877177b8bb9859aaed2f9588600b48b3f Mon Sep 17 00:00:00 2001
From: Mark H Weaver <m...@netris.org>
Date: Fri, 28 Oct 2011 13:18:43 -0400
Subject: [PATCH 1/4] Fix typos in comments of arch/mips/include/asm/inst.h
---
arch/mips/include/asm/inst.h | 4 ++--
1 files changed, 2 insertions(+), 2 deletions(-)
diff --git a/arch/mips/include/asm/inst.h b/arch/mips/include/asm/inst.h
index 7ebfc39..ab84064 100644
--- a/arch/mips/include/asm/inst.h
+++ b/arch/mips/include/asm/inst.h
@@ -251,7 +251,7 @@ struct f_format { /* FPU register format */
unsigned int func : 6;
};
-struct ma_format { /* FPU multipy and add format (MIPS IV) */
+struct ma_format { /* FPU multiply and add format (MIPS IV) */
unsigned int opcode : 6;
unsigned int fr : 5;
unsigned int ft : 5;
@@ -324,7 +324,7 @@ struct f_format { /* FPU register format */
unsigned int opcode : 6;
};
-struct ma_format { /* FPU multipy and add format (MIPS IV) */
+struct ma_format { /* FPU multiply and add format (MIPS IV) */
unsigned int fmt : 2;
unsigned int func : 4;
unsigned int fd : 5;
--
1.7.5.4
>From bf55ef4e3c2f622ac013f196affbd11b67b59223 Mon Sep 17 00:00:00 2001
From: Mark H Weaver <m...@netris.org>
Date: Fri, 28 Oct 2011 13:24:37 -0400
Subject: [PATCH 2/4] Fix handling of prefx instruction in mips/math-emu
* The instruction is named prefx, not pfetch, and its function
field is 0x17, not 0x07.
* Recognize the prefx instruction regardless of what bits happen to be
in bits 21-25, which is the format field of the floating-point ops,
but holds the base register of the prefx instruction.
---
arch/mips/include/asm/inst.h | 4 ++--
arch/mips/math-emu/cp1emu.c | 16 +++++++---------
2 files changed, 9 insertions(+), 11 deletions(-)
diff --git a/arch/mips/include/asm/inst.h b/arch/mips/include/asm/inst.h
index ab84064..3048edc 100644
--- a/arch/mips/include/asm/inst.h
+++ b/arch/mips/include/asm/inst.h
@@ -161,8 +161,8 @@ enum cop1_sdw_func {
*/
enum cop1x_func {
lwxc1_op = 0x00, ldxc1_op = 0x01,
- pfetch_op = 0x07, swxc1_op = 0x08,
- sdxc1_op = 0x09, madd_s_op = 0x20,
+ swxc1_op = 0x08, sdxc1_op = 0x09,
+ prefx_op = 0x17, madd_s_op = 0x20,
madd_d_op = 0x21, madd_e_op = 0x22,
msub_s_op = 0x28, msub_d_op = 0x29,
msub_e_op = 0x2a, nmadd_s_op = 0x30,
diff --git a/arch/mips/math-emu/cp1emu.c b/arch/mips/math-emu/cp1emu.c
index dbf2f93..87ddba1 100644
--- a/arch/mips/math-emu/cp1emu.c
+++ b/arch/mips/math-emu/cp1emu.c
@@ -739,7 +739,7 @@ static int fpux_emu(struct pt_regs *xcp, struct mips_fpu_struct *ctx,
break;
default:
- return SIGILL;
+ goto SIGILL_unless_prefx_op;
}
break;
}
@@ -809,19 +809,17 @@ static int fpux_emu(struct pt_regs *xcp, struct mips_fpu_struct *ctx,
goto copcsr;
default:
- return SIGILL;
+ goto SIGILL_unless_prefx_op;
}
break;
}
- case 0x7: /* 7 */
- if (MIPSInst_FUNC(ir) != pfetch_op) {
- return SIGILL;
- }
- /* ignore prefx operation */
- break;
-
default:
+ SIGILL_unless_prefx_op:
+ if (MIPSInst_FUNC(ir) == prefx_op) {
+ /* ignore prefx operation */
+ break;
+ }
return SIGILL;
}
--
1.7.5.4
>From 97a564e3eddbfb84844b8eccb3bd751c71dfb3eb Mon Sep 17 00:00:00 2001
From: Mark H Weaver <m...@netris.org>
Date: Fri, 28 Oct 2011 13:35:27 -0400
Subject: [PATCH 3/4] Don't process empty cause flags after simple fp move on
mips
---
arch/mips/math-emu/cp1emu.c | 4 ++--
1 files changed, 2 insertions(+), 2 deletions(-)
diff --git a/arch/mips/math-emu/cp1emu.c b/arch/mips/math-emu/cp1emu.c
index 87ddba1..fefcba2 100644
--- a/arch/mips/math-emu/cp1emu.c
+++ b/arch/mips/math-emu/cp1emu.c
@@ -912,7 +912,7 @@ static int fpu_emu(struct pt_regs *xcp, struct mips_fpu_struct *ctx,
case fmov_op:
/* an easy one */
SPFROMREG(rv.s, MIPSInst_FS(ir));
- goto copcsr;
+ break;
/* binary op on handler */
scopbop:
@@ -1099,7 +1099,7 @@ static int fpu_emu(struct pt_regs *xcp, struct mips_fpu_struct *ctx,
case fmov_op:
/* an easy one */
DPFROMREG(rv.d, MIPSInst_FS(ir));
- goto copcsr;
+ break;
/* binary op on handler */
dcopbop:{
--
1.7.5.4
>From 4051727b3007ef3675e7258ed86fa8517f86d929 Mon Sep 17 00:00:00 2001
From: Mark H Weaver <m...@netris.org>
Date: Fri, 28 Oct 2011 13:39:10 -0400
Subject: [PATCH 4/4] Support Loongson2f floating-point instructions in
mips/math-emu
* (arch/mips/include/asm/inst.h): Add Loongson2f function field values
for madd/msub/nmadd/nmsub that use the spec2 opcode, and the
Loongson2f/MIPS-5 format field value for paired-single
floating-point operations.
* (arch/mips/math-emu/cp1emu.c): Add support for the Loongson2f
instructions for madd/msub/nmadd/nmsub, which use the spec2 opcode.
Also add support for the Loongson2f instructions that use the
paired-single floating-point format.
---
arch/mips/include/asm/inst.h | 4 +-
arch/mips/math-emu/cp1emu.c | 287 +++++++++++++++++++++++++++++++++++++++++-
2 files changed, 289 insertions(+), 2 deletions(-)
diff --git a/arch/mips/include/asm/inst.h b/arch/mips/include/asm/inst.h
index 3048edc..0e8ba7c 100644
--- a/arch/mips/include/asm/inst.h
+++ b/arch/mips/include/asm/inst.h
@@ -61,6 +61,8 @@ enum spec_op {
enum spec2_op {
madd_op, maddu_op, mul_op, spec2_3_unused_op,
msub_op, msubu_op, /* more unused ops */
+ loongson_madd_op = 0x18, loongson_msub_op,
+ loongson_nmadd_op, loongson_nmsub_op,
clz_op = 0x20, clo_op,
dclz_op = 0x24, dclo_op,
sdbpp_op = 0x3f
@@ -133,7 +135,7 @@ enum cop0_com_func {
*/
enum cop1_fmt {
s_fmt, d_fmt, e_fmt, q_fmt,
- w_fmt, l_fmt
+ w_fmt, l_fmt, ps_fmt
};
/*
diff --git a/arch/mips/math-emu/cp1emu.c b/arch/mips/math-emu/cp1emu.c
index fefcba2..166b2a4 100644
--- a/arch/mips/math-emu/cp1emu.c
+++ b/arch/mips/math-emu/cp1emu.c
@@ -7,6 +7,9 @@
* Kevin D. Kissell, kev...@mips.com and Carsten Langgaard, carst...@mips.com
* Copyright (C) 2000 MIPS Technologies, Inc.
*
+ * Loongson instruction support
+ * Copyright (C) 2011 Mark H Weaver <m...@netris.org>
+ *
* This program is free software; you can distribute it and/or modify it
* under the terms of the GNU General Public License (Version 2) as
* published by the Free Software Foundation.
@@ -57,6 +60,14 @@
#endif
#define __mips 4
+#ifdef __loongson_fp
+#undef __loongson_fp
+#endif
+#if __mips >= 4 && __mips != 32
+/* Include support for Loongson floating point instructions */
+#define __loongson_fp 1
+#endif
+
/* Function which emulates a floating point instruction. */
static int fpu_emu(struct pt_regs *, struct mips_fpu_struct *,
@@ -66,6 +77,10 @@ static int fpu_emu(struct pt_regs *, struct mips_fpu_struct *,
static int fpux_emu(struct pt_regs *,
struct mips_fpu_struct *, mips_instruction, void *__user *);
#endif
+#ifdef __loongson_fp
+static int loongson_spec2_emu(struct pt_regs *,
+ struct mips_fpu_struct *, mips_instruction, void *__user *);
+#endif
/* Further private data for which no space exists in mips_fpu_struct */
@@ -203,6 +218,14 @@ static inline int cop1_64bit(struct pt_regs *xcp)
#define DPFROMREG(dp, x) DIFROMREG((dp).bits, x)
#define DPTOREG(dp, x) DITOREG((dp).bits, x)
+/* Support for Loongson paired single floating-point format */
+#define PSIFROMREG(si1, si2, x) ({ u64 di; DIFROMREG(di, x); \
+ (si1) = (u32)di; (si2) = (u32)(di >> 32); })
+#define PSITOREG(si1, si2, x) DITOREG((si1) | ((u64)(si2) << 32), x)
+
+#define PSPFROMREG(sp1, sp2, x) PSIFROMREG((sp1).bits, (sp2).bits, x)
+#define PSPTOREG(sp1, sp2, x) PSITOREG((sp1).bits, (sp2).bits, x)
+
/*
* Emulate the single floating point instruction pointed at by EPC.
* Two instructions if the instruction is in a branch delay slot.
@@ -568,6 +591,15 @@ static int cop1Emulate(struct pt_regs *xcp, struct mips_fpu_struct *ctx,
break;
#endif
+#ifdef __loongson_fp
+ case spec2_op:{
+ int sig = loongson_spec2_emu(xcp, ctx, ir, fault_addr);
+ if (sig)
+ return sig;
+ break;
+ }
+#endif
+
default:
return SIGILL;
}
@@ -646,6 +678,172 @@ DEF3OP(msub, dp, ieee754dp_mul, ieee754dp_sub, );
DEF3OP(nmadd, dp, ieee754dp_mul, ieee754dp_add, ieee754dp_neg);
DEF3OP(nmsub, dp, ieee754dp_mul, ieee754dp_sub, ieee754dp_neg);
+#ifdef __loongson_fp
+static int loongson_spec2_emu(struct pt_regs *xcp, struct mips_fpu_struct *ctx,
+ mips_instruction ir, void *__user *fault_addr)
+{
+ int rfmt; /* resulting format */
+ unsigned rcsr = 0; /* resulting csr */
+ union {
+ ieee754dp d;
+ struct {
+ ieee754sp s;
+ ieee754sp s2;
+ };
+ } rv; /* resulting value */
+
+ /* XXX maybe add a counter for loongson spec2 fp instructions? */
+ /* MIPS_FPU_EMU_INC_STATS(cp1xops); */
+
+ switch (rfmt = (MIPSInst_FFMT(ir) & 0xf)) {
+ case s_fmt:{
+ ieee754sp(*handler) (ieee754sp, ieee754sp, ieee754sp);
+ ieee754sp fd, fs, ft;
+
+ switch (MIPSInst_FUNC(ir)) {
+ case loongson_madd_op:
+ handler = fpemu_sp_madd;
+ goto scoptop;
+ case loongson_msub_op:
+ handler = fpemu_sp_msub;
+ goto scoptop;
+ case loongson_nmadd_op:
+ handler = fpemu_sp_nmadd;
+ goto scoptop;
+ case loongson_nmsub_op:
+ handler = fpemu_sp_nmsub;
+ goto scoptop;
+
+ scoptop:
+ SPFROMREG(fd, MIPSInst_FD(ir));
+ SPFROMREG(fs, MIPSInst_FS(ir));
+ SPFROMREG(ft, MIPSInst_FT(ir));
+ rv.s = (*handler) (fd, fs, ft);
+
+ copcsr:
+ if (ieee754_cxtest(IEEE754_INEXACT))
+ rcsr |= FPU_CSR_INE_X | FPU_CSR_INE_S;
+ if (ieee754_cxtest(IEEE754_UNDERFLOW))
+ rcsr |= FPU_CSR_UDF_X | FPU_CSR_UDF_S;
+ if (ieee754_cxtest(IEEE754_OVERFLOW))
+ rcsr |= FPU_CSR_OVF_X | FPU_CSR_OVF_S;
+ if (ieee754_cxtest(IEEE754_INVALID_OPERATION))
+ rcsr |= FPU_CSR_INV_X | FPU_CSR_INV_S;
+
+ break;
+
+ default:
+ return SIGILL;
+ }
+ break;
+ }
+
+ case d_fmt:{
+ ieee754dp(*handler) (ieee754dp, ieee754dp, ieee754dp);
+ ieee754dp fd, fs, ft;
+
+ switch (MIPSInst_FUNC(ir)) {
+ case loongson_madd_op:
+ handler = fpemu_dp_madd;
+ goto dcoptop;
+ case loongson_msub_op:
+ handler = fpemu_dp_msub;
+ goto dcoptop;
+ case loongson_nmadd_op:
+ handler = fpemu_dp_nmadd;
+ goto dcoptop;
+ case loongson_nmsub_op:
+ handler = fpemu_dp_nmsub;
+ goto dcoptop;
+
+ dcoptop:
+ DPFROMREG(fd, MIPSInst_FD(ir));
+ DPFROMREG(fs, MIPSInst_FS(ir));
+ DPFROMREG(ft, MIPSInst_FT(ir));
+ rv.d = (*handler) (fd, fs, ft);
+ goto copcsr;
+
+ default:
+ return SIGILL;
+ }
+ break;
+ }
+
+ case ps_fmt:{
+ ieee754sp(*handler) (ieee754sp, ieee754sp, ieee754sp);
+ struct _ieee754_csr ieee754_csr_save;
+ ieee754sp fd1, fs1, ft1;
+ ieee754sp fd2, fs2, ft2;
+
+ switch (MIPSInst_FUNC(ir)) {
+ case loongson_madd_op:
+ handler = fpemu_sp_madd;
+ goto pscoptop;
+ case loongson_msub_op:
+ handler = fpemu_sp_msub;
+ goto pscoptop;
+ case loongson_nmadd_op:
+ handler = fpemu_sp_nmadd;
+ goto pscoptop;
+ case loongson_nmsub_op:
+ handler = fpemu_sp_nmsub;
+ goto pscoptop;
+
+ pscoptop:
+ PSPFROMREG(fd1, fd2, MIPSInst_FD(ir));
+ PSPFROMREG(fs1, fs2, MIPSInst_FS(ir));
+ PSPFROMREG(ft1, ft2, MIPSInst_FT(ir));
+ rv.s = (*handler) (fd1, fs1, ft1);
+ ieee754_csr_save = ieee754_csr;
+ rv.s2 = (*handler) (fd2, fs2, ft2);
+ ieee754_csr.cx |= ieee754_csr_save.cx;
+ ieee754_csr.sx |= ieee754_csr_save.sx;
+ goto copcsr;
+
+ default:
+ return SIGILL;
+ }
+ break;
+ }
+
+ default:
+ return SIGILL;
+ }
+
+ /*
+ * Update the fpu CSR register for this operation.
+ * If an exception is required, generate a tidy SIGFPE exception,
+ * without updating the result register.
+ * Note: cause exception bits do not accumulate, they are rewritten
+ * for each op; only the flag/sticky bits accumulate.
+ */
+ ctx->fcr31 = (ctx->fcr31 & ~FPU_CSR_ALL_X) | rcsr;
+ if ((ctx->fcr31 >> 5) & ctx->fcr31 & FPU_CSR_ALL_E) {
+ /*printk ("SIGFPE: fpu csr = %08x\n",ctx->fcr31); */
+ return SIGFPE;
+ }
+
+ /*
+ * Now we can safely write the result back to the register file.
+ */
+ switch (rfmt) {
+ case d_fmt:
+ DPTOREG(rv.d, MIPSInst_FD(ir));
+ break;
+ case s_fmt:
+ SPTOREG(rv.s, MIPSInst_FD(ir));
+ break;
+ case ps_fmt:
+ PSPTOREG(rv.s, rv.s2, MIPSInst_FD(ir));
+ break;
+ default:
+ return SIGILL;
+ }
+
+ return 0;
+}
+#endif
+
static int fpux_emu(struct pt_regs *xcp, struct mips_fpu_struct *ctx,
mips_instruction ir, void *__user *fault_addr)
{
@@ -840,7 +1038,12 @@ static int fpu_emu(struct pt_regs *xcp, struct mips_fpu_struct *ctx,
unsigned cond;
union {
ieee754dp d;
- ieee754sp s;
+ struct {
+ ieee754sp s;
+#ifdef __loongson_fp
+ ieee754sp s2; /* for Loongson paired singles */
+#endif
+ };
int w;
#ifdef __mips64
s64 l;
@@ -1210,6 +1413,83 @@ static int fpu_emu(struct pt_regs *xcp, struct mips_fpu_struct *ctx,
break;
}
+#ifdef __loongson_fp
+ case ps_fmt:{ /* 6 */
+ /* Support for Loongson paired single fp instructions */
+ union {
+ ieee754sp(*b) (ieee754sp, ieee754sp);
+ ieee754sp(*u) (ieee754sp);
+ } handler;
+
+ switch (MIPSInst_FUNC(ir)) {
+ /* binary ops */
+ case fadd_op:
+ handler.b = ieee754sp_add;
+ goto pscopbop;
+ case fsub_op:
+ handler.b = ieee754sp_sub;
+ goto pscopbop;
+ case fmul_op:
+ handler.b = ieee754sp_mul;
+ goto pscopbop;
+
+ /* unary ops */
+ case fabs_op:
+ handler.u = ieee754sp_abs;
+ goto pscopuop;
+ case fneg_op:
+ handler.u = ieee754sp_neg;
+ goto pscopuop;
+ case fmov_op:
+ /* an easy one */
+ PSPFROMREG(rv.s, rv.s2, MIPSInst_FS(ir));
+ break;
+
+ pscopbop: /* paired binary op handler */
+ {
+ struct _ieee754_csr ieee754_csr_save;
+ ieee754sp fs1, ft1;
+ ieee754sp fs2, ft2;
+
+ PSPFROMREG(fs1, fs2, MIPSInst_FS(ir));
+ PSPFROMREG(ft1, ft2, MIPSInst_FT(ir));
+ rv.s = (*handler.b) (fs1, ft1);
+ ieee754_csr_save = ieee754_csr;
+ rv.s2 = (*handler.b) (fs2, ft2);
+ ieee754_csr.cx |= ieee754_csr_save.cx;
+ ieee754_csr.sx |= ieee754_csr_save.sx;
+ goto copcsr;
+ }
+ pscopuop: /* paired unary op handler */
+ {
+ struct _ieee754_csr ieee754_csr_save;
+ ieee754sp fs1;
+ ieee754sp fs2;
+
+ PSPFROMREG(fs1, fs2, MIPSInst_FS(ir));
+ rv.s = (*handler.u) (fs1);
+ ieee754_csr_save = ieee754_csr;
+ rv.s2 = (*handler.u) (fs2);
+ ieee754_csr.cx |= ieee754_csr_save.cx;
+ ieee754_csr.sx |= ieee754_csr_save.sx;
+ goto copcsr;
+ }
+ break;
+
+ default:
+ if (MIPSInst_FUNC(ir) >= fcmp_op) {
+ /* Loongson fp hardware handles all
+ cases of fp compare insns, so we
+ shouldn't have to */
+ printk ("Loongson paired-single fp compare"
+ " unimplemented in cp1emu.c\n");
+ }
+ return SIGILL;
+ }
+ break;
+ }
+#endif
+
case w_fmt:{
ieee754sp fs;
@@ -1299,6 +1579,11 @@ static int fpu_emu(struct pt_regs *xcp, struct mips_fpu_struct *ctx,
DITOREG(rv.l, MIPSInst_FD(ir));
break;
#endif
+#ifdef __loongson_fp
+ case ps_fmt:
+ PSPTOREG(rv.s, rv.s2, MIPSInst_FD(ir));
+ break;
+#endif
default:
return SIGILL;
}
--
1.7.5.4
--- binutils-2.21/opcodes/mips-opc.c.orig 2010-10-28 09:49:51.000000000 -0400
+++ binutils-2.21/opcodes/mips-opc.c 2011-10-28 15:09:23.000000000 -0400
@@ -830,7 +830,7 @@
{"madd.s", "D,S,T", 0x72000018, 0xffe0003f, RD_S|RD_T|WR_D|FP_S, 0, IL2F },
{"madd.ps", "D,R,S,T", 0x4c000026, 0xfc00003f, RD_R|RD_S|RD_T|WR_D|FP_D, 0, I5_33 },
{"madd.ps", "D,S,T", 0x45600018, 0xffe0003f, RD_S|RD_T|WR_D|FP_D, 0, IL2E },
-{"madd.ps", "D,S,T", 0x71600018, 0xffe0003f, RD_S|RD_T|WR_D|FP_D, 0, IL2F },
+{"madd.ps", "D,S,T", 0x72c00018, 0xffe0003f, RD_S|RD_T|WR_D|FP_D, 0, IL2F },
{"madd", "s,t", 0x0000001c, 0xfc00ffff, RD_s|RD_t|WR_HILO, 0, L1 },
{"madd", "s,t", 0x70000000, 0xfc00ffff, RD_s|RD_t|MOD_HILO, 0, I32|N55 },
{"madd", "s,t", 0x70000000, 0xfc00ffff, RD_s|RD_t|WR_HILO|IS_M, 0, G1 },
@@ -940,7 +940,7 @@
{"msub.s", "D,S,T", 0x72000019, 0xffe0003f, RD_S|RD_T|WR_D|FP_S, 0, IL2F },
{"msub.ps", "D,R,S,T", 0x4c00002e, 0xfc00003f, RD_R|RD_S|RD_T|WR_D|FP_D, 0, I5_33 },
{"msub.ps", "D,S,T", 0x45600019, 0xffe0003f, RD_S|RD_T|WR_D|FP_D, 0, IL2E },
-{"msub.ps", "D,S,T", 0x71600019, 0xffe0003f, RD_S|RD_T|WR_D|FP_D, 0, IL2F },
+{"msub.ps", "D,S,T", 0x72c00019, 0xffe0003f, RD_S|RD_T|WR_D|FP_D, 0, IL2F },
{"msub", "s,t", 0x0000001e, 0xfc00ffff, RD_s|RD_t|WR_HILO, 0, L1 },
{"msub", "s,t", 0x70000004, 0xfc00ffff, RD_s|RD_t|MOD_HILO, 0, I32|N55 },
{"msub", "7,s,t", 0x70000004, 0xfc00e7ff, MOD_a|RD_s|RD_t, 0, D32 },
@@ -1056,7 +1056,7 @@
{"nmadd.s", "D,S,T", 0x7200001a, 0xffe0003f, RD_S|RD_T|WR_D|FP_S, 0, IL2F },
{"nmadd.ps","D,R,S,T", 0x4c000036, 0xfc00003f, RD_R|RD_S|RD_T|WR_D|FP_D, 0, I5_33 },
{"nmadd.ps", "D,S,T", 0x4560001a, 0xffe0003f, RD_S|RD_T|WR_D|FP_D, 0, IL2E },
-{"nmadd.ps", "D,S,T", 0x7160001a, 0xffe0003f, RD_S|RD_T|WR_D|FP_D, 0, IL2F },
+{"nmadd.ps", "D,S,T", 0x72c0001a, 0xffe0003f, RD_S|RD_T|WR_D|FP_D, 0, IL2F },
{"nmsub.d", "D,R,S,T", 0x4c000039, 0xfc00003f, RD_R|RD_S|RD_T|WR_D|FP_D, 0, I4_33 },
{"nmsub.d", "D,S,T", 0x4620001b, 0xffe0003f, RD_S|RD_T|WR_D|FP_D, 0, IL2E },
{"nmsub.d", "D,S,T", 0x7220001b, 0xffe0003f, RD_S|RD_T|WR_D|FP_D, 0, IL2F },
@@ -1065,7 +1065,7 @@
{"nmsub.s", "D,S,T", 0x7200001b, 0xffe0003f, RD_S|RD_T|WR_D|FP_S, 0, IL2F },
{"nmsub.ps","D,R,S,T", 0x4c00003e, 0xfc00003f, RD_R|RD_S|RD_T|WR_D|FP_D, 0, I5_33 },
{"nmsub.ps", "D,S,T", 0x4560001b, 0xffe0003f, RD_S|RD_T|WR_D|FP_D, 0, IL2E },
-{"nmsub.ps", "D,S,T", 0x7160001b, 0xffe0003f, RD_S|RD_T|WR_D|FP_D, 0, IL2F },
+{"nmsub.ps", "D,S,T", 0x72c0001b, 0xffe0003f, RD_S|RD_T|WR_D|FP_D, 0, IL2F },
/* nop is at the start of the table. */
{"nor", "d,v,t", 0x00000027, 0xfc0007ff, WR_d|RD_s|RD_t, 0, I1 },
{"nor", "t,r,I", 0, (int) M_NOR_I, INSN_MACRO, 0, I1 },
