Linux-MIPS Archive on lore.kernel.org
 help / Atom feed
* [PATCH] MIPS: math-emu: Add madd/msub/nmadd/nmsub emulation for Loongson-3
@ 2019-02-06 12:03 Huacai Chen
  0 siblings, 0 replies; 1+ messages in thread
From: Huacai Chen @ 2019-02-06 12:03 UTC (permalink / raw)
  To: Paul Burton, Ralf Baechle, James Hogan
  Cc: linux-mips, linux-mips, Fuxin Zhang, Zhangjin Wu, Huacai Chen,
	Huacai Chen, Pei Huang

Add madd.s/madd.d/msub.s/msub.d/nmadd.s/nmadd.d/nmsub.s/nmsub.d
emulation for Loongson-3. MIPS R2 suggest these instructions be
unfused, but Loongson-3 suggest these instructions be fused, which
is similar to maddf/msubf in MIPS R6.

Signed-off-by: Huacai Chen <chenhc@lemote.com>
Signed-off-by: Pei Huang <huangpei@loongson.cn>
---
 arch/mips/math-emu/cp1emu.c     | 32 +++++++++++++++++++++++++
 arch/mips/math-emu/dp_maddf.c   | 53 +++++++++++++++++++++++++++--------------
 arch/mips/math-emu/ieee754.h    | 16 +++++++++++++
 arch/mips/math-emu/ieee754int.h |  1 +
 arch/mips/math-emu/sp_maddf.c   | 53 +++++++++++++++++++++++++++--------------
 5 files changed, 119 insertions(+), 36 deletions(-)

diff --git a/arch/mips/math-emu/cp1emu.c b/arch/mips/math-emu/cp1emu.c
index e60e290..2f99ec2 100644
--- a/arch/mips/math-emu/cp1emu.c
+++ b/arch/mips/math-emu/cp1emu.c
@@ -1463,6 +1463,7 @@ static union ieee754sp fpemu_sp_rsqrt(union ieee754sp s)
 	return ieee754sp_div(ieee754sp_one(0), ieee754sp_sqrt(s));
 }
 
+#ifndef CONFIG_CPU_LOONGSON3
 DEF3OP(madd, sp, ieee754sp_mul, ieee754sp_add, );
 DEF3OP(msub, sp, ieee754sp_mul, ieee754sp_sub, );
 DEF3OP(nmadd, sp, ieee754sp_mul, ieee754sp_add, ieee754sp_neg);
@@ -1471,6 +1472,7 @@ DEF3OP(madd, dp, ieee754dp_mul, ieee754dp_add, );
 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);
+#endif
 
 static int fpux_emu(struct pt_regs *xcp, struct mips_fpu_struct *ctx,
 	mips_instruction ir, void __user **fault_addr)
@@ -1525,6 +1527,20 @@ static int fpux_emu(struct pt_regs *xcp, struct mips_fpu_struct *ctx,
 			}
 			break;
 
+#ifdef CONFIG_CPU_LOONGSON3
+		case madd_s_op:
+			handler = ieee754sp_madd;
+			goto scoptop;
+		case msub_s_op:
+			handler = ieee754sp_msub;
+			goto scoptop;
+		case nmadd_s_op:
+			handler = ieee754sp_nmadd;
+			goto scoptop;
+		case nmsub_s_op:
+			handler = ieee754sp_nmsub;
+			goto scoptop;
+#else
 		case madd_s_op:
 			handler = fpemu_sp_madd;
 			goto scoptop;
@@ -1537,6 +1553,7 @@ static int fpux_emu(struct pt_regs *xcp, struct mips_fpu_struct *ctx,
 		case nmsub_s_op:
 			handler = fpemu_sp_nmsub;
 			goto scoptop;
+#endif
 
 		      scoptop:
 			SPFROMREG(fr, MIPSInst_FR(ir));
@@ -1621,6 +1638,20 @@ static int fpux_emu(struct pt_regs *xcp, struct mips_fpu_struct *ctx,
 			}
 			break;
 
+#ifdef CONFIG_CPU_LOONGSON3
+		case madd_d_op:
+			handler = ieee754dp_madd;
+			goto dcoptop;
+		case msub_d_op:
+			handler = ieee754dp_msub;
+			goto dcoptop;
+		case nmadd_d_op:
+			handler = ieee754dp_nmadd;
+			goto dcoptop;
+		case nmsub_d_op:
+			handler = ieee754dp_nmsub;
+			goto dcoptop;
+#else
 		case madd_d_op:
 			handler = fpemu_dp_madd;
 			goto dcoptop;
@@ -1633,6 +1664,7 @@ static int fpux_emu(struct pt_regs *xcp, struct mips_fpu_struct *ctx,
 		case nmsub_d_op:
 			handler = fpemu_dp_nmsub;
 			goto dcoptop;
+#endif
 
 		      dcoptop:
 			DPFROMREG(fr, MIPSInst_FR(ir));
diff --git a/arch/mips/math-emu/dp_maddf.c b/arch/mips/math-emu/dp_maddf.c
index 7ea2f82..a9c2db4 100644
--- a/arch/mips/math-emu/dp_maddf.c
+++ b/arch/mips/math-emu/dp_maddf.c
@@ -71,6 +71,12 @@ static union ieee754dp _dp_maddf(union ieee754dp z, union ieee754dp x,
 
 	ieee754_clearcx();
 
+	rs = xs ^ ys;
+	if (flags & MADDF_NEGATE_PRODUCT)
+		rs ^= 1;
+	if (flags & MADDF_NEGATE_ADDITION)
+		zs ^= 1;
+
 	/*
 	 * Handle the cases when at least one of x, y or z is a NaN.
 	 * Order of precedence is sNaN, qNaN and z, x, y.
@@ -107,9 +113,7 @@ static union ieee754dp _dp_maddf(union ieee754dp z, union ieee754dp x,
 	case CLPAIR(IEEE754_CLASS_INF, IEEE754_CLASS_NORM):
 	case CLPAIR(IEEE754_CLASS_INF, IEEE754_CLASS_DNORM):
 	case CLPAIR(IEEE754_CLASS_INF, IEEE754_CLASS_INF):
-		if ((zc == IEEE754_CLASS_INF) &&
-		    ((!(flags & MADDF_NEGATE_PRODUCT) && (zs != (xs ^ ys))) ||
-		     ((flags & MADDF_NEGATE_PRODUCT) && (zs == (xs ^ ys))))) {
+		if ((zc == IEEE754_CLASS_INF) && (zs != rs)) {
 			/*
 			 * Cases of addition of infinities with opposite signs
 			 * or subtraction of infinities with same signs.
@@ -119,15 +123,10 @@ static union ieee754dp _dp_maddf(union ieee754dp z, union ieee754dp x,
 		}
 		/*
 		 * z is here either not an infinity, or an infinity having the
-		 * same sign as product (x*y) (in case of MADDF.D instruction)
-		 * or product -(x*y) (in MSUBF.D case). The result must be an
-		 * infinity, and its sign is determined only by the value of
-		 * (flags & MADDF_NEGATE_PRODUCT) and the signs of x and y.
+		 * same sign as product (x*y). The result must be an infinity,
+		 * and its sign is determined only by the sign of product (x*y).
 		 */
-		if (flags & MADDF_NEGATE_PRODUCT)
-			return ieee754dp_inf(1 ^ (xs ^ ys));
-		else
-			return ieee754dp_inf(xs ^ ys);
+		return ieee754dp_inf(rs);
 
 	case CLPAIR(IEEE754_CLASS_ZERO, IEEE754_CLASS_ZERO):
 	case CLPAIR(IEEE754_CLASS_ZERO, IEEE754_CLASS_NORM):
@@ -138,10 +137,7 @@ static union ieee754dp _dp_maddf(union ieee754dp z, union ieee754dp x,
 			return ieee754dp_inf(zs);
 		if (zc == IEEE754_CLASS_ZERO) {
 			/* Handle cases +0 + (-0) and similar ones. */
-			if ((!(flags & MADDF_NEGATE_PRODUCT)
-					&& (zs == (xs ^ ys))) ||
-			    ((flags & MADDF_NEGATE_PRODUCT)
-					&& (zs != (xs ^ ys))))
+			if (zs == rs)
 				/*
 				 * Cases of addition of zeros of equal signs
 				 * or subtraction of zeroes of opposite signs.
@@ -190,9 +186,6 @@ static union ieee754dp _dp_maddf(union ieee754dp z, union ieee754dp x,
 	assert(ym & DP_HIDDEN_BIT);
 
 	re = xe + ye;
-	rs = xs ^ ys;
-	if (flags & MADDF_NEGATE_PRODUCT)
-		rs ^= 1;
 
 	/* shunt to top of word */
 	xm <<= 64 - (DP_FBITS + 1);
@@ -343,3 +336,27 @@ union ieee754dp ieee754dp_msubf(union ieee754dp z, union ieee754dp x,
 {
 	return _dp_maddf(z, x, y, MADDF_NEGATE_PRODUCT);
 }
+
+union ieee754dp ieee754dp_madd(union ieee754dp z, union ieee754dp x,
+				union ieee754dp y)
+{
+	return _dp_maddf(z, x, y, 0);
+}
+
+union ieee754dp ieee754dp_msub(union ieee754dp z, union ieee754dp x,
+				union ieee754dp y)
+{
+	return _dp_maddf(z, x, y, MADDF_NEGATE_ADDITION);
+}
+
+union ieee754dp ieee754dp_nmadd(union ieee754dp z, union ieee754dp x,
+				union ieee754dp y)
+{
+	return _dp_maddf(z, x, y, MADDF_NEGATE_PRODUCT|MADDF_NEGATE_ADDITION);
+}
+
+union ieee754dp ieee754dp_nmsub(union ieee754dp z, union ieee754dp x,
+				union ieee754dp y)
+{
+	return _dp_maddf(z, x, y, MADDF_NEGATE_PRODUCT);
+}
diff --git a/arch/mips/math-emu/ieee754.h b/arch/mips/math-emu/ieee754.h
index e0eb7a9..e12bb9f 100644
--- a/arch/mips/math-emu/ieee754.h
+++ b/arch/mips/math-emu/ieee754.h
@@ -80,6 +80,14 @@ union ieee754sp ieee754sp_maddf(union ieee754sp z, union ieee754sp x,
 				union ieee754sp y);
 union ieee754sp ieee754sp_msubf(union ieee754sp z, union ieee754sp x,
 				union ieee754sp y);
+union ieee754sp ieee754sp_madd(union ieee754sp z, union ieee754sp x,
+				union ieee754sp y);
+union ieee754sp ieee754sp_msub(union ieee754sp z, union ieee754sp x,
+				union ieee754sp y);
+union ieee754sp ieee754sp_nmadd(union ieee754sp z, union ieee754sp x,
+				union ieee754sp y);
+union ieee754sp ieee754sp_nmsub(union ieee754sp z, union ieee754sp x,
+				union ieee754sp y);
 int ieee754sp_2008class(union ieee754sp x);
 union ieee754sp ieee754sp_fmin(union ieee754sp x, union ieee754sp y);
 union ieee754sp ieee754sp_fmina(union ieee754sp x, union ieee754sp y);
@@ -115,6 +123,14 @@ union ieee754dp ieee754dp_maddf(union ieee754dp z, union ieee754dp x,
 				union ieee754dp y);
 union ieee754dp ieee754dp_msubf(union ieee754dp z, union ieee754dp x,
 				union ieee754dp y);
+union ieee754dp ieee754dp_madd(union ieee754dp z, union ieee754dp x,
+				union ieee754dp y);
+union ieee754dp ieee754dp_msub(union ieee754dp z, union ieee754dp x,
+				union ieee754dp y);
+union ieee754dp ieee754dp_nmadd(union ieee754dp z, union ieee754dp x,
+				union ieee754dp y);
+union ieee754dp ieee754dp_nmsub(union ieee754dp z, union ieee754dp x,
+				union ieee754dp y);
 int ieee754dp_2008class(union ieee754dp x);
 union ieee754dp ieee754dp_fmin(union ieee754dp x, union ieee754dp y);
 union ieee754dp ieee754dp_fmina(union ieee754dp x, union ieee754dp y);
diff --git a/arch/mips/math-emu/ieee754int.h b/arch/mips/math-emu/ieee754int.h
index 06ac0e2..302815c 100644
--- a/arch/mips/math-emu/ieee754int.h
+++ b/arch/mips/math-emu/ieee754int.h
@@ -28,6 +28,7 @@
 
 enum maddf_flags {
 	MADDF_NEGATE_PRODUCT	= 1 << 0,
+	MADDF_NEGATE_ADDITION	= 1 << 1,
 };
 
 static inline void ieee754_clearcx(void)
diff --git a/arch/mips/math-emu/sp_maddf.c b/arch/mips/math-emu/sp_maddf.c
index 07ba675..d334181 100644
--- a/arch/mips/math-emu/sp_maddf.c
+++ b/arch/mips/math-emu/sp_maddf.c
@@ -39,6 +39,12 @@ static union ieee754sp _sp_maddf(union ieee754sp z, union ieee754sp x,
 
 	ieee754_clearcx();
 
+	rs = xs ^ ys;
+	if (flags & MADDF_NEGATE_PRODUCT)
+		rs ^= 1;
+	if (flags & MADDF_NEGATE_ADDITION)
+		zs ^= 1;
+
 	/*
 	 * Handle the cases when at least one of x, y or z is a NaN.
 	 * Order of precedence is sNaN, qNaN and z, x, y.
@@ -76,9 +82,7 @@ static union ieee754sp _sp_maddf(union ieee754sp z, union ieee754sp x,
 	case CLPAIR(IEEE754_CLASS_INF, IEEE754_CLASS_NORM):
 	case CLPAIR(IEEE754_CLASS_INF, IEEE754_CLASS_DNORM):
 	case CLPAIR(IEEE754_CLASS_INF, IEEE754_CLASS_INF):
-		if ((zc == IEEE754_CLASS_INF) &&
-		    ((!(flags & MADDF_NEGATE_PRODUCT) && (zs != (xs ^ ys))) ||
-		     ((flags & MADDF_NEGATE_PRODUCT) && (zs == (xs ^ ys))))) {
+		if ((zc == IEEE754_CLASS_INF) && (zs != rs)) {
 			/*
 			 * Cases of addition of infinities with opposite signs
 			 * or subtraction of infinities with same signs.
@@ -88,15 +92,10 @@ static union ieee754sp _sp_maddf(union ieee754sp z, union ieee754sp x,
 		}
 		/*
 		 * z is here either not an infinity, or an infinity having the
-		 * same sign as product (x*y) (in case of MADDF.D instruction)
-		 * or product -(x*y) (in MSUBF.D case). The result must be an
-		 * infinity, and its sign is determined only by the value of
-		 * (flags & MADDF_NEGATE_PRODUCT) and the signs of x and y.
+		 * same sign as product (x*y). The result must be an infinity,
+		 * and its sign is determined only by the sign of product (x*y).
 		 */
-		if (flags & MADDF_NEGATE_PRODUCT)
-			return ieee754sp_inf(1 ^ (xs ^ ys));
-		else
-			return ieee754sp_inf(xs ^ ys);
+		return ieee754sp_inf(rs);
 
 	case CLPAIR(IEEE754_CLASS_ZERO, IEEE754_CLASS_ZERO):
 	case CLPAIR(IEEE754_CLASS_ZERO, IEEE754_CLASS_NORM):
@@ -107,10 +106,7 @@ static union ieee754sp _sp_maddf(union ieee754sp z, union ieee754sp x,
 			return ieee754sp_inf(zs);
 		if (zc == IEEE754_CLASS_ZERO) {
 			/* Handle cases +0 + (-0) and similar ones. */
-			if ((!(flags & MADDF_NEGATE_PRODUCT)
-					&& (zs == (xs ^ ys))) ||
-			    ((flags & MADDF_NEGATE_PRODUCT)
-					&& (zs != (xs ^ ys))))
+			if (zs == rs)
 				/*
 				 * Cases of addition of zeros of equal signs
 				 * or subtraction of zeroes of opposite signs.
@@ -161,9 +157,6 @@ static union ieee754sp _sp_maddf(union ieee754sp z, union ieee754sp x,
 	assert(ym & SP_HIDDEN_BIT);
 
 	re = xe + ye;
-	rs = xs ^ ys;
-	if (flags & MADDF_NEGATE_PRODUCT)
-		rs ^= 1;
 
 	/* Multiple 24 bit xm and ym to give 48 bit results */
 	rm64 = (uint64_t)xm * ym;
@@ -263,3 +256,27 @@ union ieee754sp ieee754sp_msubf(union ieee754sp z, union ieee754sp x,
 {
 	return _sp_maddf(z, x, y, MADDF_NEGATE_PRODUCT);
 }
+
+union ieee754sp ieee754sp_madd(union ieee754sp z, union ieee754sp x,
+				union ieee754sp y)
+{
+	return _sp_maddf(z, x, y, 0);
+}
+
+union ieee754sp ieee754sp_msub(union ieee754sp z, union ieee754sp x,
+				union ieee754sp y)
+{
+	return _sp_maddf(z, x, y, MADDF_NEGATE_ADDITION);
+}
+
+union ieee754sp ieee754sp_nmadd(union ieee754sp z, union ieee754sp x,
+				union ieee754sp y)
+{
+	return _sp_maddf(z, x, y, MADDF_NEGATE_PRODUCT|MADDF_NEGATE_ADDITION);
+}
+
+union ieee754sp ieee754sp_nmsub(union ieee754sp z, union ieee754sp x,
+				union ieee754sp y)
+{
+	return _sp_maddf(z, x, y, MADDF_NEGATE_PRODUCT);
+}
-- 
2.7.0


^ permalink raw reply	[flat|nested] 1+ messages in thread

only message in thread, back to index

Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2019-02-06 12:03 [PATCH] MIPS: math-emu: Add madd/msub/nmadd/nmsub emulation for Loongson-3 Huacai Chen

Linux-MIPS Archive on lore.kernel.org

Archives are clonable:
	git clone --mirror https://lore.kernel.org/linux-mips/0 linux-mips/git/0.git

	# If you have public-inbox 1.1+ installed, you may
	# initialize and index your mirror using the following commands:
	public-inbox-init -V2 linux-mips linux-mips/ https://lore.kernel.org/linux-mips \
		linux-mips@vger.kernel.org linux-mips@archiver.kernel.org
	public-inbox-index linux-mips


Newsgroup available over NNTP:
	nntp://nntp.lore.kernel.org/org.kernel.vger.linux-mips


AGPL code for this site: git clone https://public-inbox.org/ public-inbox