Merge pull request #1980 from flintlib/nfloat8

Specialize addmul, submul for nfloat
flintlib · May 17, 2024 · 7528e42 · 7528e42
2 parents 8611346 + 9840dfe
commit 7528e42
Show file tree

Hide file tree

Showing 8 changed files with 496 additions and 7 deletions.
diff --git a/doc/source/nfloat.rst b/doc/source/nfloat.rst
@@ -252,6 +252,8 @@ These methods are interchangeable with their ``gr`` counterparts.
               int nfloat_add(nfloat_ptr res, nfloat_srcptr x, nfloat_srcptr y, gr_ctx_t ctx)
               int nfloat_sub(nfloat_ptr res, nfloat_srcptr x, nfloat_srcptr y, gr_ctx_t ctx)
               int nfloat_mul(nfloat_ptr res, nfloat_srcptr x, nfloat_srcptr y, gr_ctx_t ctx)
+              int nfloat_submul(nfloat_ptr res, nfloat_srcptr x, nfloat_srcptr y, gr_ctx_t ctx)
+              int nfloat_addmul(nfloat_ptr res, nfloat_srcptr x, nfloat_srcptr y, gr_ctx_t ctx)
 
 .. function:: int nfloat_mul_2exp_si(nfloat_ptr res, nfloat_srcptr x, slong y, gr_ctx_t ctx)
 
@@ -303,6 +305,8 @@ code for reduced overhead.
               int _nfloat_vec_sub(nfloat_ptr res, nfloat_srcptr x, nfloat_srcptr y, slong len, gr_ctx_t ctx)
               int _nfloat_vec_mul(nfloat_ptr res, nfloat_srcptr x, nfloat_srcptr y, slong len, gr_ctx_t ctx)
               int _nfloat_vec_mul_scalar(nfloat_ptr res, nfloat_srcptr x, slong len, nfloat_srcptr y, gr_ctx_t ctx)
+              int _nfloat_vec_addmul_scalar(nfloat_ptr res, nfloat_srcptr x, slong len, nfloat_srcptr y, gr_ctx_t ctx)
+              int _nfloat_vec_submul_scalar(nfloat_ptr res, nfloat_srcptr x, slong len, nfloat_srcptr y, gr_ctx_t ctx)
 
 .. function:: int _nfloat_vec_dot(nfloat_ptr res, nfloat_srcptr initial, int subtract, nfloat_srcptr x, nfloat_srcptr y, slong len, gr_ctx_t ctx)
               int _nfloat_vec_dot_rev(nfloat_ptr res, nfloat_srcptr initial, int subtract, nfloat_srcptr x, nfloat_srcptr y, slong len, gr_ctx_t ctx)

diff --git a/src/nfloat.h b/src/nfloat.h
@@ -303,6 +303,8 @@ int _nfloat_sub_n(nfloat_ptr res, nn_srcptr xd, slong xexp, int xsgnbit, nn_srcp
 int nfloat_add(nfloat_ptr res, nfloat_srcptr x, nfloat_srcptr y, gr_ctx_t ctx);
 int nfloat_sub(nfloat_ptr res, nfloat_srcptr x, nfloat_srcptr y, gr_ctx_t ctx);
 int nfloat_mul(nfloat_ptr res, nfloat_srcptr x, nfloat_srcptr y, gr_ctx_t ctx);
+int nfloat_addmul(nfloat_ptr res, nfloat_srcptr x, nfloat_srcptr y, gr_ctx_t ctx);
+int nfloat_submul(nfloat_ptr res, nfloat_srcptr x, nfloat_srcptr y, gr_ctx_t ctx);
 
 int nfloat_mul_2exp_si(nfloat_ptr res, nfloat_srcptr x, slong y, gr_ctx_t ctx);
 
@@ -344,6 +346,8 @@ int _nfloat_vec_add(nfloat_ptr res, nfloat_srcptr x, nfloat_srcptr y, slong len,
 int _nfloat_vec_sub(nfloat_ptr res, nfloat_srcptr x, nfloat_srcptr y, slong len, gr_ctx_t ctx);
 int _nfloat_vec_mul(nfloat_ptr res, nfloat_srcptr x, nfloat_srcptr y, slong len, gr_ctx_t ctx);
 int _nfloat_vec_mul_scalar(nfloat_ptr res, nfloat_srcptr x, slong len, nfloat_srcptr y, gr_ctx_t ctx);
+int _nfloat_vec_addmul_scalar(nfloat_ptr res, nfloat_srcptr x, slong len, nfloat_srcptr y, gr_ctx_t ctx);
+int _nfloat_vec_submul_scalar(nfloat_ptr res, nfloat_srcptr x, slong len, nfloat_srcptr y, gr_ctx_t ctx);
 
 int _nfloat_vec_dot(nfloat_ptr res, nfloat_srcptr initial, int subtract, nfloat_srcptr x, nfloat_srcptr y, slong len, gr_ctx_t ctx);
 int _nfloat_vec_dot_rev(nfloat_ptr res, nfloat_srcptr initial, int subtract, nfloat_srcptr x, nfloat_srcptr y, slong len, gr_ctx_t ctx);

diff --git a/src/nfloat/ctx.c b/src/nfloat/ctx.c
@@ -87,8 +87,10 @@ gr_method_tab_input _nfloat_methods_input[] =
     {GR_METHOD_MUL_SI,          (gr_funcptr) nfloat_mul_si},
     {GR_METHOD_MUL_FMPZ,        (gr_funcptr) nfloat_mul_fmpz},
     {GR_METHOD_MUL_TWO,         (gr_funcptr) nfloat_mul_two},
+*/
     {GR_METHOD_ADDMUL,          (gr_funcptr) nfloat_addmul},
     {GR_METHOD_SUBMUL,          (gr_funcptr) nfloat_submul},
+/*
     {GR_METHOD_SQR,             (gr_funcptr) nfloat_sqr},
 */
     {GR_METHOD_DIV,             (gr_funcptr) nfloat_div},
@@ -163,6 +165,8 @@ gr_method_tab_input _nfloat_methods_input[] =
     {GR_METHOD_VEC_SUB,                 (gr_funcptr) _nfloat_vec_sub},
     {GR_METHOD_VEC_MUL,                 (gr_funcptr) _nfloat_vec_mul},
     {GR_METHOD_VEC_MUL_SCALAR,          (gr_funcptr) _nfloat_vec_mul_scalar},
+    {GR_METHOD_VEC_ADDMUL_SCALAR,          (gr_funcptr) _nfloat_vec_addmul_scalar},
+    {GR_METHOD_VEC_SUBMUL_SCALAR,          (gr_funcptr) _nfloat_vec_submul_scalar},
     {GR_METHOD_VEC_DOT,         (gr_funcptr) _nfloat_vec_dot},
     {GR_METHOD_VEC_DOT_REV,     (gr_funcptr) _nfloat_vec_dot_rev},
 /*