From d876f5327985eac3bf3109e9429febc8a8954ff5 Mon Sep 17 00:00:00 2001 From: Ulrich Drepper Date: Wed, 29 Dec 1999 18:11:48 +0000 Subject: Update. 1999-12-29 Ulrich Drepper * soft-fp/*: Tons of new files to implement floating-point arithmetic in software. Contributed by Richard Henderson, Jakub Jelinek and others. --- soft-fp/sysdeps/alpha/sfp-machine.h | 35 +++++++++++++++++++++++++++++++++++ 1 file changed, 35 insertions(+) create mode 100644 soft-fp/sysdeps/alpha/sfp-machine.h (limited to 'soft-fp/sysdeps/alpha/sfp-machine.h') diff --git a/soft-fp/sysdeps/alpha/sfp-machine.h b/soft-fp/sysdeps/alpha/sfp-machine.h new file mode 100644 index 0000000000..73b934ddce --- /dev/null +++ b/soft-fp/sysdeps/alpha/sfp-machine.h @@ -0,0 +1,35 @@ +#define _FP_W_TYPE_SIZE 64 +#define _FP_W_TYPE unsigned long +#define _FP_WS_TYPE signed long +#define _FP_I_TYPE long + +#define _alpha_mul_64_128(rhi, rlo, x, y) \ + __asm__("umulh %2,%3,%0; mulq %2,%3,%1" \ + : "=&r"(rhi), "=r"(rlo) : "r"(x), "r"(y)) + +#define _FP_MUL_MEAT_S(R,X,Y) \ + _FP_MUL_MEAT_1_imm(_FP_WFRACBITS_S,R,X,Y) +#define _FP_MUL_MEAT_D(R,X,Y) \ + _FP_MUL_MEAT_1_wide(_FP_WFRACBITS_D,R,X,Y,_alpha_mul_64_128) +#define _FP_MUL_MEAT_Q(R,X,Y) \ + _FP_MUL_MEAT_2_wide(_FP_WFRACBITS_Q,R,X,Y,_alpha_mul_64_128) + +#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(D,R,X,Y) +#define _FP_DIV_MEAT_Q(R,X,Y) _FP_DIV_MEAT_2_udiv(Q,R,X,Y) + +#define _FP_NANFRAC_S _FP_QNANBIT_S +#define _FP_NANFRAC_D _FP_QNANBIT_D +#define _FP_NANFRAC_Q _FP_QNANBIT_Q, 0 +/* FIXME: This is just a wild guess */ +#define _FP_NANSIGN_S 1 +#define _FP_NANSIGN_D 1 +#define _FP_NANSIGN_Q 1 + +#define _FP_KEEPNANFRACP 1 +#define _FP_CHOOSENAN(fs, wc, R, X, Y, OP) \ + do { \ + R##_s = Y##_s; \ + _FP_FRAC_COPY_##wc(R,Y); \ + R##_c = FP_CLS_NAN; \ + } while (0) -- cgit v1.2.3