summaryrefslogtreecommitdiffstats
path: root/src
diff options
context:
space:
mode:
authorQuincey Koziol <koziol@hdfgroup.org>2003-10-18 16:27:25 (GMT)
committerQuincey Koziol <koziol@hdfgroup.org>2003-10-18 16:27:25 (GMT)
commitfea52b8d91c36e167ca7fc50f3bffda988170c83 (patch)
treef9f045fd5a1b01461909ae3131b7a4bf42c88a32 /src
parente6aed98f8f5ee24737995571914deb1e20331f36 (diff)
downloadhdf5-fea52b8d91c36e167ca7fc50f3bffda988170c83.zip
hdf5-fea52b8d91c36e167ca7fc50f3bffda988170c83.tar.gz
hdf5-fea52b8d91c36e167ca7fc50f3bffda988170c83.tar.bz2
[svn-r7664] Purpose:
More refactoring Description: Re-wrote inner loop of integer conversion routines with "psuedo-template" macros, to hoist 'if-else's out of inner loop in an easy to understand way. This speeds up the integer type conversions by 15-20%. Platforms tested: FreeBSD 4.9 (sleipnir) too minor to require h5committest
Diffstat (limited to 'src')
-rw-r--r--src/H5Tconv.c199
1 files changed, 124 insertions, 75 deletions
diff --git a/src/H5Tconv.c b/src/H5Tconv.c
index 278917c..b70eae4 100644
--- a/src/H5Tconv.c
+++ b/src/H5Tconv.c
@@ -149,8 +149,8 @@ H5FL_BLK_DEFINE_STATIC(array_seq);
* which allows the core conversion macro to be invoked as necessary.
*
*/
-#define H5T_CONV_sS_CORE(STYPE,DTYPE,ST,DT,D_MIN,D_MAX) { \
- *((DT*)d) = (DT)(*((ST*)s)); \
+#define H5T_CONV_sS_CORE(S,D,STYPE,DTYPE,ST,DT,D_MIN,D_MAX) { \
+ *((DT*)D) = (DT)(*((ST*)S)); \
}
#define H5T_CONV_sS(STYPE,DTYPE,ST,DT,D_MIN,D_MAX) { \
@@ -158,14 +158,14 @@ H5FL_BLK_DEFINE_STATIC(array_seq);
H5T_CONV(H5T_CONV_sS, STYPE, DTYPE, ST, DT, D_MIN, D_MAX, nelmts-1) \
}
-#define H5T_CONV_sU_CORE(STYPE,DTYPE,ST,DT,D_MIN,D_MAX) { \
- if (*((ST*)s)<0) { \
+#define H5T_CONV_sU_CORE(S,D,STYPE,DTYPE,ST,DT,D_MIN,D_MAX) { \
+ if (*((ST*)S)<0) { \
if (!H5T_overflow_g || \
- (H5T_overflow_g)(src_id, dst_id, s, d)<0) { \
- *((DT*)d) = 0; \
+ (H5T_overflow_g)(src_id, dst_id, S, D)<0) { \
+ *((DT*)D) = 0; \
} \
} else { \
- *((DT*)d) = (DT)(*((ST*)s)); \
+ *((DT*)D) = (DT)(*((ST*)S)); \
} \
}
@@ -174,14 +174,14 @@ H5FL_BLK_DEFINE_STATIC(array_seq);
H5T_CONV(H5T_CONV_sU, STYPE, DTYPE, ST, DT, D_MIN, D_MAX, nelmts-1) \
}
-#define H5T_CONV_uS_CORE(STYPE,DTYPE,ST,DT,D_MIN,D_MAX) { \
- if (*((ST*)s) > (D_MAX)) { \
+#define H5T_CONV_uS_CORE(S,D,STYPE,DTYPE,ST,DT,D_MIN,D_MAX) { \
+ if (*((ST*)S) > (D_MAX)) { \
if (!H5T_overflow_g || \
- (H5T_overflow_g)(src_id, dst_id, s, d)<0) { \
- *((DT*)d) = (D_MAX); \
+ (H5T_overflow_g)(src_id, dst_id, S, D)<0) { \
+ *((DT*)D) = (D_MAX); \
} \
} else { \
- *((DT*)d) = (DT)(*((ST*)s)); \
+ *((DT*)D) = (DT)(*((ST*)S)); \
} \
}
@@ -190,8 +190,8 @@ H5FL_BLK_DEFINE_STATIC(array_seq);
H5T_CONV(H5T_CONV_uS, STYPE, DTYPE, ST, DT, D_MIN, D_MAX, nelmts-1) \
}
-#define H5T_CONV_uU_CORE(STYPE,DTYPE,ST,DT,D_MIN,D_MAX) { \
- *((DT*)d) = (DT)(*((ST*)s)); \
+#define H5T_CONV_uU_CORE(S,D,STYPE,DTYPE,ST,DT,D_MIN,D_MAX) { \
+ *((DT*)D) = (DT)(*((ST*)S)); \
}
#define H5T_CONV_uU(STYPE,DTYPE,ST,DT,D_MIN,D_MAX) { \
@@ -199,19 +199,19 @@ H5FL_BLK_DEFINE_STATIC(array_seq);
H5T_CONV(H5T_CONV_uU, STYPE, DTYPE, ST, DT, D_MIN, D_MAX, nelmts-1) \
}
-#define H5T_CONV_Ss_CORE(STYPE,DTYPE,ST,DT,D_MIN,D_MAX) { \
- if (*((ST*)s) > (DT)(D_MAX)) { \
+#define H5T_CONV_Ss_CORE(S,D,STYPE,DTYPE,ST,DT,D_MIN,D_MAX) { \
+ if (*((ST*)S) > (DT)(D_MAX)) { \
if (!H5T_overflow_g || \
- (H5T_overflow_g)(src_id, dst_id, s, d)<0) { \
- *((DT*)d) = (D_MAX); \
+ (H5T_overflow_g)(src_id, dst_id, S, D)<0) { \
+ *((DT*)D) = (D_MAX); \
} \
- } else if (*((ST*)s) < (D_MIN)) { \
+ } else if (*((ST*)S) < (D_MIN)) { \
if (!H5T_overflow_g || \
- (H5T_overflow_g)(src_id, dst_id, s, d)<0) { \
- *((DT*)d) = (D_MIN); \
+ (H5T_overflow_g)(src_id, dst_id, S, D)<0) { \
+ *((DT*)D) = (D_MIN); \
} \
} else { \
- *((DT*)d) = (DT)(*((ST*)s)); \
+ *((DT*)D) = (DT)(*((ST*)S)); \
} \
}
@@ -220,20 +220,20 @@ H5FL_BLK_DEFINE_STATIC(array_seq);
H5T_CONV(H5T_CONV_Ss, STYPE, DTYPE, ST, DT, D_MIN, D_MAX, 0) \
}
-#define H5T_CONV_Su_CORE(STYPE,DTYPE,ST,DT,D_MIN,D_MAX) { \
- if (*((ST*)s) < 0) { \
+#define H5T_CONV_Su_CORE(S,D,STYPE,DTYPE,ST,DT,D_MIN,D_MAX) { \
+ if (*((ST*)S) < 0) { \
if (!H5T_overflow_g || \
- (H5T_overflow_g)(src_id, dst_id, s, d)<0) { \
- *((DT*)d) = 0; \
+ (H5T_overflow_g)(src_id, dst_id, S, D)<0) { \
+ *((DT*)D) = 0; \
} \
- } else if (sizeof(ST)>sizeof(DT) && *((ST*)s)>(ST)(D_MAX)) { \
+ } else if (sizeof(ST)>sizeof(DT) && *((ST*)S)>(ST)(D_MAX)) { \
/*sign vs. unsign ok in previous line*/ \
if (!H5T_overflow_g || \
- (H5T_overflow_g)(src_id, dst_id, s, d)<0) { \
- *((DT*)d) = (D_MAX); \
+ (H5T_overflow_g)(src_id, dst_id, S, D)<0) { \
+ *((DT*)D) = (D_MAX); \
} \
} else { \
- *((DT*)d) = (DT)(*((ST*)s)); \
+ *((DT*)D) = (DT)(*((ST*)S)); \
} \
}
@@ -242,14 +242,14 @@ H5FL_BLK_DEFINE_STATIC(array_seq);
H5T_CONV(H5T_CONV_Su, STYPE, DTYPE, ST, DT, D_MIN, D_MAX, 0) \
}
-#define H5T_CONV_Us_CORE(STYPE,DTYPE,ST,DT,D_MIN,D_MAX) { \
- if (*((ST*)s) > (D_MAX)) { \
+#define H5T_CONV_Us_CORE(S,D,STYPE,DTYPE,ST,DT,D_MIN,D_MAX) { \
+ if (*((ST*)S) > (D_MAX)) { \
if (!H5T_overflow_g || \
- (H5T_overflow_g)(src_id, dst_id, s, d)<0) { \
- *((DT*)d) = (D_MAX); \
+ (H5T_overflow_g)(src_id, dst_id, S, D)<0) { \
+ *((DT*)D) = (D_MAX); \
} \
} else { \
- *((DT*)d) = (DT)(*((ST*)s)); \
+ *((DT*)D) = (DT)(*((ST*)S)); \
} \
}
@@ -258,14 +258,14 @@ H5FL_BLK_DEFINE_STATIC(array_seq);
H5T_CONV(H5T_CONV_Us, STYPE, DTYPE, ST, DT, D_MIN, D_MAX, 0) \
}
-#define H5T_CONV_Uu_CORE(STYPE,DTYPE,ST,DT,D_MIN,D_MAX) { \
- if (*((ST*)s) > (D_MAX)) { \
+#define H5T_CONV_Uu_CORE(S,D,STYPE,DTYPE,ST,DT,D_MIN,D_MAX) { \
+ if (*((ST*)S) > (D_MAX)) { \
if (!H5T_overflow_g || \
- (H5T_overflow_g)(src_id, dst_id, s, d)<0) { \
- *((DT*)d) = (D_MAX); \
+ (H5T_overflow_g)(src_id, dst_id, S, D)<0) { \
+ *((DT*)D) = (D_MAX); \
} \
} else { \
- *((DT*)d) = (DT)(*((ST*)s)); \
+ *((DT*)D) = (DT)(*((ST*)S)); \
} \
}
@@ -274,14 +274,14 @@ H5FL_BLK_DEFINE_STATIC(array_seq);
H5T_CONV(H5T_CONV_Uu, STYPE, DTYPE, ST, DT, D_MIN, D_MAX, 0) \
}
-#define H5T_CONV_su_CORE(STYPE,DTYPE,ST,DT,D_MIN,D_MAX) { \
- if (*((ST*)s) < 0) { \
+#define H5T_CONV_su_CORE(S,D,STYPE,DTYPE,ST,DT,D_MIN,D_MAX) { \
+ if (*((ST*)S) < 0) { \
if (!H5T_overflow_g || \
- (H5T_overflow_g)(src_id, dst_id, s, d)<0) { \
- *((DT*)d) = 0; \
+ (H5T_overflow_g)(src_id, dst_id, S, D)<0) { \
+ *((DT*)D) = 0; \
} \
} else { \
- *((DT*)d) = (DT)(*((ST*)s)); \
+ *((DT*)D) = (DT)(*((ST*)S)); \
} \
}
@@ -290,14 +290,14 @@ H5FL_BLK_DEFINE_STATIC(array_seq);
H5T_CONV(H5T_CONV_su, STYPE, DTYPE, ST, DT, D_MIN, D_MAX, 0) \
}
-#define H5T_CONV_us_CORE(STYPE,DTYPE,ST,DT,D_MIN,D_MAX) { \
- if (*((ST*)s) > (D_MAX)) { \
+#define H5T_CONV_us_CORE(S,D,STYPE,DTYPE,ST,DT,D_MIN,D_MAX) { \
+ if (*((ST*)S) > (D_MAX)) { \
if (!H5T_overflow_g || \
- (H5T_overflow_g)(src_id, dst_id, s, d)<0) { \
- *((DT*)d) = (D_MAX); \
+ (H5T_overflow_g)(src_id, dst_id, S, D)<0) { \
+ *((DT*)D) = (D_MAX); \
} \
} else { \
- *((DT*)d) = (DT)(*((ST*)s)); \
+ *((DT*)D) = (DT)(*((ST*)S)); \
} \
}
@@ -314,7 +314,6 @@ H5FL_BLK_DEFINE_STATIC(array_seq);
H5T_t *st, *dt; /*data type descriptors */ \
long_long aligned; /*largest integer type, aligned */ \
hbool_t s_mv, d_mv; /*move data to align it? */ \
- size_t dt_size=sizeof(DT); /*needed by CI_END macro */ \
size_t s_stride, d_stride; /*src and dst strides */ \
int direction; /*1=left-to-right, -1=rt-to-lt */ \
\
@@ -371,30 +370,21 @@ H5FL_BLK_DEFINE_STATIC(array_seq);
CI_INC_SRC(s_mv) \
CI_INC_DST(d_mv) \
\
- for (elmtno=0; elmtno<nelmts; elmtno++) { \
- /* Alignment */ \
- if (s_mv) { \
- HDmemcpy(&aligned, src, sizeof(ST)); \
- s = (uint8_t*)&aligned; \
- } else { \
- s = src; \
- } \
- if (d_mv) { \
- d = (uint8_t*)&aligned; \
- } else { \
- d = dst; \
- } \
- \
- /* ... user-defined stuff here -- the conversion ... */ \
- H5_GLUE(GUTS,_CORE)(STYPE,DTYPE,ST,DT,D_MIN,D_MAX) \
- \
- /* Copy destination to final location */ \
- if (d_mv) HDmemcpy(dst, &aligned, dt_size); \
- \
- /* Advance pointers */ \
- src = (char *)src + direction * s_stride; \
- dst = (char *)dst + direction * d_stride; \
- } \
+ if (s_mv && d_mv) { \
+ /* Alignment is required for both source and dest */ \
+ s = (uint8_t*)&aligned; \
+ H5T_CONV_LOOP(PRE_SALIGN,PRE_DALIGN,POST_SALIGN,POST_DALIGN,GUTS,s,d,STYPE,DTYPE,ST,DT,D_MIN,D_MAX) \
+ } else if(s_mv) { \
+ /* Alignment is required only for source */ \
+ s = (uint8_t*)&aligned; \
+ H5T_CONV_LOOP(PRE_SALIGN,PRE_DNOALIGN,POST_SALIGN,POST_DNOALIGN,GUTS,s,dst,STYPE,DTYPE,ST,DT,D_MIN,D_MAX) \
+ } else if(d_mv) { \
+ /* Alignment is required only for destination */ \
+ H5T_CONV_LOOP(PRE_SNOALIGN,PRE_DALIGN,POST_SNOALIGN,POST_DALIGN,GUTS,src,d,STYPE,DTYPE,ST,DT,D_MIN,D_MAX) \
+ } else { \
+ /* Alignment is not required for both source and destination */ \
+ H5T_CONV_LOOP(PRE_SNOALIGN,PRE_DNOALIGN,POST_SNOALIGN,POST_DNOALIGN,GUTS,src,dst,STYPE,DTYPE,ST,DT,D_MIN,D_MAX) \
+ } \
break; \
\
default: \
@@ -403,6 +393,65 @@ H5FL_BLK_DEFINE_STATIC(array_seq);
} \
}
+/* Macro defining action on source data which needs to be aligned (before main action) */
+#define H5T_CONV_LOOP_PRE_SALIGN(ST) { \
+ HDmemcpy(&aligned, src, sizeof(ST)); \
+}
+
+/* Macro defining action on source data which doesn't need to be aligned (before main action) */
+#define H5T_CONV_LOOP_PRE_SNOALIGN(ST) { \
+}
+
+/* Macro defining action on destination data which needs to be aligned (before main action) */
+#define H5T_CONV_LOOP_PRE_DALIGN(DT) { \
+ d = (uint8_t*)&aligned; \
+}
+
+/* Macro defining action on destination data which doesn't need to be aligned (before main action) */
+#define H5T_CONV_LOOP_PRE_DNOALIGN(DT) { \
+}
+
+/* Macro defining action on source data which needs to be aligned (after main action) */
+#define H5T_CONV_LOOP_POST_SALIGN(ST) { \
+}
+
+/* Macro defining action on source data which doesn't need to be aligned (after main action) */
+#define H5T_CONV_LOOP_POST_SNOALIGN(ST) { \
+}
+
+/* Macro defining action on destination data which needs to be aligned (after main action) */
+#define H5T_CONV_LOOP_POST_DALIGN(DT) { \
+ HDmemcpy(dst, &aligned, sizeof(DT)); \
+}
+
+/* Macro defining action on destination data which doesn't need to be aligned (after main action) */
+#define H5T_CONV_LOOP_POST_DNOALIGN(DT) { \
+}
+
+/* The inner loop of the type conversion macro */
+#define H5T_CONV_LOOP(PRE_SALIGN_GUTS,PRE_DALIGN_GUTS,POST_SALIGN_GUTS,POST_DALIGN_GUTS,GUTS,S,D,STYPE,DTYPE,ST,DT,D_MIN,D_MAX) \
+ for (elmtno=0; elmtno<nelmts; elmtno++) { \
+ /* Handle source pre-alignment */ \
+ H5_GLUE(H5T_CONV_LOOP_,PRE_SALIGN_GUTS)(ST) \
+ \
+ /* Handle destination pre-alignment */ \
+ H5_GLUE(H5T_CONV_LOOP_,PRE_DALIGN_GUTS)(DT) \
+ \
+ /* ... user-defined stuff here -- the conversion ... */ \
+ H5_GLUE(GUTS,_CORE)(S,D,STYPE,DTYPE,ST,DT,D_MIN,D_MAX) \
+ \
+ /* Handle source post-alignment */ \
+ H5_GLUE(H5T_CONV_LOOP_,POST_SALIGN_GUTS)(ST) \
+ \
+ /* Handle destination post-alignment */ \
+ H5_GLUE(H5T_CONV_LOOP_,POST_DALIGN_GUTS)(DT) \
+ \
+ /* Advance pointers */ \
+ src = (char *)src + direction * s_stride; \
+ dst = (char *)dst + direction * d_stride; \
+ }
+
+
#ifdef H5T_DEBUG
/* Print alignment statistics */