Implement fma in soft-fp.

8 files changed, 32 insertions, 12 deletions

diff --git a/ports/sysdeps/mips/ieee754/s_fma.c b/ports/sysdeps/mips/ieee754/s_fma.c
new file mode 100644
index 0000000..5741414
--- /dev/null
+++ b/ports/sysdeps/mips/ieee754/s_fma.c
@@ -0,0 +1,5 @@
+#ifdef __mips_hard_float
+# include <sysdeps/ieee754/dbl-64/s_fma.c>
+#else
+# include <soft-fp/fmadf4.c>
+#endif
diff --git a/ports/sysdeps/mips/ieee754/s_fmaf.c b/ports/sysdeps/mips/ieee754/s_fmaf.c
new file mode 100644
index 0000000..30bcdae
--- /dev/null
+++ b/ports/sysdeps/mips/ieee754/s_fmaf.c
@@ -0,0 +1,5 @@
+#ifdef __mips_hard_float
+# include <sysdeps/ieee754/dbl-64/s_fmaf.c>
+#else
+# include <soft-fp/fmasf4.c>
+#endif
diff --git a/ports/sysdeps/mips/ieee754/s_fmal.c b/ports/sysdeps/mips/ieee754/s_fmal.c
new file mode 100644
index 0000000..6b83e91
--- /dev/null
+++ b/ports/sysdeps/mips/ieee754/s_fmal.c
@@ -0,0 +1,7 @@
+#include <sgidefs.h>
+
+#if _MIPS_SIM == _ABIO32
+# error "long double fma being compiled for o32 ABI"
+#endif
+
+#include <soft-fp/fmatf4.c>
diff --git a/ports/sysdeps/mips/mips32/Implies b/ports/sysdeps/mips/mips32/Implies
index 6473f25..42df98f 100644
--- a/ports/sysdeps/mips/mips32/Implies
+++ b/ports/sysdeps/mips/mips32/Implies
@@ -1,3 +1,4 @@
 mips/ieee754
+mips/soft-fp
 mips
 wordsize-32
diff --git a/ports/sysdeps/mips/mips64/n32/s_fma.c b/ports/sysdeps/mips/mips64/n32/s_fma.c
deleted file mode 100644
index 74a1e01..0000000
--- a/ports/sysdeps/mips/mips64/n32/s_fma.c
+++ /dev/null
@@ -1,6 +0,0 @@
-/* MIPS long double is implemented in software by fp-bit (as of GCC
-   4.7) without support for exceptions or rounding modes, so the fma
-   implementation in terms of long double is slow and will not produce
-   correctly rounding results.  */
-
-#include <sysdeps/ieee754/dbl-64/s_fma.c>
diff --git a/ports/sysdeps/mips/mips64/n64/s_fma.c b/ports/sysdeps/mips/mips64/n64/s_fma.c
deleted file mode 100644
index 74a1e01..0000000
--- a/ports/sysdeps/mips/mips64/n64/s_fma.c
+++ /dev/null
@@ -1,6 +0,0 @@
-/* MIPS long double is implemented in software by fp-bit (as of GCC
-   4.7) without support for exceptions or rounding modes, so the fma
-   implementation in terms of long double is slow and will not produce
-   correctly rounding results.  */
-
-#include <sysdeps/ieee754/dbl-64/s_fma.c>
diff --git a/ports/sysdeps/mips/mips64/soft-fp/sfp-machine.h b/ports/sysdeps/mips/mips64/soft-fp/sfp-machine.h
index 1bdde5a..9cfd6fb 100644
--- a/ports/sysdeps/mips/mips64/soft-fp/sfp-machine.h
+++ b/ports/sysdeps/mips/mips64/soft-fp/sfp-machine.h
@@ -13,6 +13,13 @@
 #define _FP_MUL_MEAT_Q(R,X,Y)					\
   _FP_MUL_MEAT_2_wide_3mul(_FP_WFRACBITS_Q,R,X,Y,umul_ppmm)
 
+#define _FP_MUL_MEAT_DW_S(R,X,Y)				\
+  _FP_MUL_MEAT_DW_1_imm(_FP_WFRACBITS_S,R,X,Y)
+#define _FP_MUL_MEAT_DW_D(R,X,Y)				\
+  _FP_MUL_MEAT_DW_1_wide(_FP_WFRACBITS_D,R,X,Y,umul_ppmm)
+#define _FP_MUL_MEAT_DW_Q(R,X,Y)				\
+  _FP_MUL_MEAT_DW_2_wide_3mul(_FP_WFRACBITS_Q,R,X,Y,umul_ppmm)
+
 #define _FP_DIV_MEAT_S(R,X,Y)	_FP_DIV_MEAT_1_imm(S,R,X,Y,_FP_DIV_HELP_imm)
 #define _FP_DIV_MEAT_D(R,X,Y)	_FP_DIV_MEAT_1_udiv_norm(D,R,X,Y)
 #define _FP_DIV_MEAT_Q(R,X,Y)	_FP_DIV_MEAT_2_udiv(Q,R,X,Y)
diff --git a/ports/sysdeps/mips/soft-fp/sfp-machine.h b/ports/sysdeps/mips/soft-fp/sfp-machine.h
index 8ccfaa6..a60bef7 100644
--- a/ports/sysdeps/mips/soft-fp/sfp-machine.h
+++ b/ports/sysdeps/mips/soft-fp/sfp-machine.h
@@ -10,6 +10,13 @@
 #define _FP_MUL_MEAT_Q(R,X,Y)				\
   _FP_MUL_MEAT_4_wide(_FP_WFRACBITS_Q,R,X,Y,umul_ppmm)
 
+#define _FP_MUL_MEAT_DW_S(R,X,Y)				\
+  _FP_MUL_MEAT_DW_1_wide(_FP_WFRACBITS_S,R,X,Y,umul_ppmm)
+#define _FP_MUL_MEAT_DW_D(R,X,Y)				\
+  _FP_MUL_MEAT_DW_2_wide(_FP_WFRACBITS_D,R,X,Y,umul_ppmm)
+#define _FP_MUL_MEAT_DW_Q(R,X,Y)				\
+  _FP_MUL_MEAT_DW_4_wide(_FP_WFRACBITS_Q,R,X,Y,umul_ppmm)
+
 #define _FP_DIV_MEAT_S(R,X,Y)	_FP_DIV_MEAT_1_udiv_norm(S,R,X,Y)
 #define _FP_DIV_MEAT_D(R,X,Y)	_FP_DIV_MEAT_2_udiv(D,R,X,Y)
 #define _FP_DIV_MEAT_Q(R,X,Y)	_FP_DIV_MEAT_4_udiv(Q,R,X,Y)