| -rw-r--r-- | lib/jit_hppa-fpu.c | 106 |
diff --git a/lib/jit_hppa-fpu.c b/lib/jit_hppa-fpu.c index d48d5ae..c676884 100644 --- a/lib/jit_hppa-fpu.c +++ b/lib/jit_hppa-fpu.c @@ -271,11 +271,19 @@ static void _f54(jit_state_t*,jit_int32_t,jit_int32_t,jit_int32_t, #define negr_f(r0,r1) FNEG_S(r1,r0) #define negr_d(r0,r1) FNEG_D(r1,r0) #define sqrtr_f(r0,r1) FSQRT_S(r1,r0) -#define fmar_f(r0,r1,r2,r3) FMPYFADD_S(r1,r2,r3,r0) -#define fmsr_f(r0,r1,r2,r3) FMPYFSUB_S(r1,r2,r3,r0) +#define fmar_f(r0,r1,r2,r3) _fmar_f(_jit,r0,r1,r2,r3) +static void _fmar_f(jit_state_t*, + jit_int32_t,jit_int32_t,jit_int32_t,jit_int32_t); +#define fmsr_f(r0,r1,r2,r3) _fmsr_f(_jit,r0,r1,r2,r3) +static void _fmsr_f(jit_state_t*, + jit_int32_t,jit_int32_t,jit_int32_t,jit_int32_t); #define sqrtr_d(r0,r1) FSQRT_D(r1,r0) -#define fmar_d(r0,r1,r2,r3) FMPYFADD_D(r1,r2,r3,r0) -#define fmsr_d(r0,r1,r2,r3) FMPYFSUB_D(r1,r2,r3,r0) +#define fmar_d(r0,r1,r2,r3) _fmar_d(_jit,r0,r1,r2,r3) +static void _fmar_d(jit_state_t*, + jit_int32_t,jit_int32_t,jit_int32_t,jit_int32_t); +#define fmsr_d(r0,r1,r2,r3) _fmsr_d(_jit,r0,r1,r2,r3) +static void _fmsr_d(jit_state_t*, + jit_int32_t,jit_int32_t,jit_int32_t,jit_int32_t); #define extr_f(r0,r1) _extr_f(_jit,r0,r1) static void _extr_f(jit_state_t*,jit_int32_t,jit_int32_t); #define extr_d(r0,r1) _extr_d(_jit,r0,r1) @@ -704,6 +712,51 @@ _truncr_d_i(jit_state_t *_jit, jit_int32_t r0, jit_int32_t r1) } static void +_fmar_f(jit_state_t *_jit, + jit_int32_t r0, jit_int32_t r1, jit_int32_t r2, jit_int32_t r3) +{ +#if 1 + jit_int32_t reg; + if (r0 != r3) { + mulr_f(r0, r1, r2); + addr_f(r0, r0, r3); + } + else { + reg = jit_get_reg(jit_class_fpr); + mulr_f(rn(reg), r1, r2); + addr_f(r0, rn(reg), r3); + jit_unget_reg(reg); + } +#else + FMPYFADD_S(r1, r2, r3, r0); +#endif +} + +static void +_fmsr_f(jit_state_t *_jit, + jit_int32_t r0, jit_int32_t r1, jit_int32_t r2, jit_int32_t r3) +{ + jit_int32_t reg; +#if 1 + if (r0 != r3) { + mulr_f(r0, r1, r2); + subr_f(r0, r0, r3); + } + else { + reg = jit_get_reg(jit_class_fpr); + mulr_f(rn(reg), r1, r2); + subr_f(r0, rn(reg), r3); + jit_unget_reg(reg); + } +#else + reg = jit_get_reg(jit_class_fpr); + negr_f(rn(reg), r3); + fmar_f(r0, r1, r2, rn(reg)); + jit_unget_reg(reg); +#endif +} + +static void _movi_f(jit_state_t *_jit, jit_int32_t r0, jit_float32_t *i0) { union { @@ -735,6 +788,51 @@ _movi_w_f(jit_state_t *_jit, jit_int32_t r0, jit_word_t i0) } static void +_fmar_d(jit_state_t *_jit, + jit_int32_t r0, jit_int32_t r1, jit_int32_t r2, jit_int32_t r3) +{ +#if 1 + jit_int32_t reg; + if (r0 != r3) { + mulr_d(r0, r1, r2); + addr_d(r0, r0, r3); + } + else { + reg = jit_get_reg(jit_class_fpr); + mulr_d(rn(reg), r1, r2); + addr_d(r0, rn(reg), r3); + jit_unget_reg(reg); + } +#else + FMPYFADD_D(r1, r2, r3, r0); +#endif +} + +static void +_fmsr_d(jit_state_t *_jit, + jit_int32_t r0, jit_int32_t r1, jit_int32_t r2, jit_int32_t r3) +{ + jit_int32_t reg; +#if 1 + if (r0 != r3) { + mulr_d(r0, r1, r2); + subr_d(r0, r0, r3); + } + else { + reg = jit_get_reg(jit_class_fpr); + mulr_d(rn(reg), r1, r2); + subr_d(r0, rn(reg), r3); + jit_unget_reg(reg); + } +#else + reg = jit_get_reg(jit_class_fpr); + negr_d(rn(reg), r3); + fmar_d(r0, r1, r2, rn(reg)); + jit_unget_reg(reg); +#endif +} + +static void _movi_d(jit_state_t *_jit, jit_int32_t r0, jit_float64_t *i0) { union { |