From 861e66848f05001fcaca4fe91dbf67cd186649dc Mon Sep 17 00:00:00 2001
From: Josh Blum <josh@joshknows.com>
Date: Fri, 7 Oct 2011 17:30:42 -0700
Subject: usrp2: work on alternative OTW formats
---
host/lib/convert/convert_common.hpp | 66 +++++++++++++++++++++++------
host/lib/convert/convert_fc32_with_sse2.cpp | 16 +++----
host/lib/convert/convert_fc64_with_sse2.cpp | 8 ++--
host/lib/convert/gen_convert_general.py | 32 +++++++++++++-
4 files changed, 96 insertions(+), 26 deletions(-)
(limited to 'host/lib/convert')
diff --git a/host/lib/convert/convert_common.hpp b/host/lib/convert/convert_common.hpp
index 612fa312b..56b718292 100644
--- a/host/lib/convert/convert_common.hpp
+++ b/host/lib/convert/convert_common.hpp
@@ -65,18 +65,18 @@ typedef boost::int8_t s8_t;
typedef boost::uint32_t item32_t;
/***********************************************************************
- * Convert complex short buffer to items32
+ * Convert complex short buffer to items32 sc16
**********************************************************************/
-static UHD_INLINE item32_t sc16_to_item32(sc16_t num, double){
+static UHD_INLINE item32_t sc16_to_item32_sc16(sc16_t num, double){
boost::uint16_t real = num.real();
boost::uint16_t imag = num.imag();
return (item32_t(real) << 16) | (item32_t(imag) << 0);
}
/***********************************************************************
- * Convert items32 buffer to complex short
+ * Convert items32 sc16 buffer to complex short
**********************************************************************/
-static UHD_INLINE sc16_t item32_to_sc16(item32_t item, double){
+static UHD_INLINE sc16_t item32_sc16_to_sc16(item32_t item, double){
return sc16_t(
boost::int16_t(item >> 16),
boost::int16_t(item >> 0)
@@ -84,18 +84,18 @@ static UHD_INLINE sc16_t item32_to_sc16(item32_t item, double){
}
/***********************************************************************
- * Convert complex float buffer to items32 (no swap)
+ * Convert complex float buffer to items32 sc16
**********************************************************************/
-static UHD_INLINE item32_t fc32_to_item32(fc32_t num, double scale_factor){
+static UHD_INLINE item32_t fc32_to_item32_sc16(fc32_t num, double scale_factor){
boost::uint16_t real = boost::int16_t(num.real()*float(scale_factor));
boost::uint16_t imag = boost::int16_t(num.imag()*float(scale_factor));
return (item32_t(real) << 16) | (item32_t(imag) << 0);
}
/***********************************************************************
- * Convert items32 buffer to complex float
+ * Convert items32 sc16 buffer to complex float
**********************************************************************/
-static UHD_INLINE fc32_t item32_to_fc32(item32_t item, double scale_factor){
+static UHD_INLINE fc32_t item32_sc16_to_fc32(item32_t item, double scale_factor){
return fc32_t(
float(boost::int16_t(item >> 16)*float(scale_factor)),
float(boost::int16_t(item >> 0)*float(scale_factor))
@@ -103,22 +103,64 @@ static UHD_INLINE fc32_t item32_to_fc32(item32_t item, double scale_factor){
}
/***********************************************************************
- * Convert complex double buffer to items32 (no swap)
+ * Convert complex double buffer to items32 sc16
**********************************************************************/
-static UHD_INLINE item32_t fc64_to_item32(fc64_t num, double scale_factor){
+static UHD_INLINE item32_t fc64_to_item32_sc16(fc64_t num, double scale_factor){
boost::uint16_t real = boost::int16_t(num.real()*scale_factor);
boost::uint16_t imag = boost::int16_t(num.imag()*scale_factor);
return (item32_t(real) << 16) | (item32_t(imag) << 0);
}
/***********************************************************************
- * Convert items32 buffer to complex double
+ * Convert items32 sc16 buffer to complex double
**********************************************************************/
-static UHD_INLINE fc64_t item32_to_fc64(item32_t item, double scale_factor){
+static UHD_INLINE fc64_t item32_sc16_to_fc64(item32_t item, double scale_factor){
return fc64_t(
float(boost::int16_t(item >> 16)*scale_factor),
float(boost::int16_t(item >> 0)*scale_factor)
);
}
+/***********************************************************************
+ * Convert items32 sc8 buffer to complex short
+ **********************************************************************/
+static UHD_INLINE void item32_sc8_to_sc16(item32_t item, sc16_t &out0, sc16_t &out1, double){
+ out0 = sc16_t(
+ boost::int8_t(item >> 8),
+ boost::int8_t(item >> 0)
+ );
+ out1 = sc16_t(
+ boost::int8_t(item >> 24),
+ boost::int8_t(item >> 16)
+ );
+}
+
+/***********************************************************************
+ * Convert items32 sc8 buffer to complex float
+ **********************************************************************/
+static UHD_INLINE void item32_sc8_to_fc32(item32_t item, fc32_t &out0, fc32_t &out1, double scale_factor){
+ out0 = fc32_t(
+ float(boost::int8_t(item >> 8)*float(scale_factor)),
+ float(boost::int8_t(item >> 0)*float(scale_factor))
+ );
+ out1 = fc32_t(
+ float(boost::int8_t(item >> 24)*float(scale_factor)),
+ float(boost::int8_t(item >> 16)*float(scale_factor))
+ );
+}
+
+/***********************************************************************
+ * Convert items32 sc8 buffer to complex double
+ **********************************************************************/
+static UHD_INLINE void item32_sc8_to_fc64(item32_t item, fc64_t &out0, fc64_t &out1, double scale_factor){
+ out0 = fc64_t(
+ float(boost::int8_t(item >> 8)*scale_factor),
+ float(boost::int8_t(item >> 0)*scale_factor)
+ );
+ out1 = fc64_t(
+ float(boost::int8_t(item >> 24)*scale_factor),
+ float(boost::int8_t(item >> 16)*scale_factor)
+ );
+}
+
#endif /* INCLUDED_LIBUHD_CONVERT_COMMON_HPP */
diff --git a/host/lib/convert/convert_fc32_with_sse2.cpp b/host/lib/convert/convert_fc32_with_sse2.cpp
index 34c85db80..b8d1aa8cc 100644
--- a/host/lib/convert/convert_fc32_with_sse2.cpp
+++ b/host/lib/convert/convert_fc32_with_sse2.cpp
@@ -51,7 +51,7 @@ DECLARE_CONVERTER(fc32, 1, sc16_item32_le, 1, PRIORITY_CUSTOM){
//dispatch according to alignment
switch (size_t(input) & 0xf){
case 0x8:
- output[i] = fc32_to_item32(input[i], float(scale_factor)); i++;
+ output[i] = fc32_to_item32_sc16(input[i], float(scale_factor)); i++;
case 0x0:
convert_fc32_1_to_item32_1_nswap_guts(_)
break;
@@ -60,7 +60,7 @@ DECLARE_CONVERTER(fc32, 1, sc16_item32_le, 1, PRIORITY_CUSTOM){
//convert remainder
for (; i < nsamps; i++){
- output[i] = fc32_to_item32(input[i], float(scale_factor));
+ output[i] = fc32_to_item32_sc16(input[i], float(scale_factor));
}
}
@@ -93,7 +93,7 @@ DECLARE_CONVERTER(fc32, 1, sc16_item32_be, 1, PRIORITY_CUSTOM){
//dispatch according to alignment
switch (size_t(input) & 0xf){
case 0x8:
- output[i] = uhd::byteswap(fc32_to_item32(input[i], float(scale_factor))); i++;
+ output[i] = uhd::byteswap(fc32_to_item32_sc16(input[i], float(scale_factor))); i++;
case 0x0:
convert_fc32_1_to_item32_1_bswap_guts(_)
break;
@@ -102,7 +102,7 @@ DECLARE_CONVERTER(fc32, 1, sc16_item32_be, 1, PRIORITY_CUSTOM){
//convert remainder
for (; i < nsamps; i++){
- output[i] = uhd::byteswap(fc32_to_item32(input[i], float(scale_factor)));
+ output[i] = uhd::byteswap(fc32_to_item32_sc16(input[i], float(scale_factor)));
}
}
@@ -138,7 +138,7 @@ DECLARE_CONVERTER(sc16_item32_le, 1, fc32, 1, PRIORITY_CUSTOM){
//dispatch according to alignment
switch (size_t(output) & 0xf){
case 0x8:
- output[i] = item32_to_fc32(input[i], float(scale_factor)); i++;
+ output[i] = item32_sc16_to_fc32(input[i], float(scale_factor)); i++;
case 0x0:
convert_item32_1_to_fc32_1_nswap_guts(_)
break;
@@ -147,7 +147,7 @@ DECLARE_CONVERTER(sc16_item32_le, 1, fc32, 1, PRIORITY_CUSTOM){
//convert remainder
for (; i < nsamps; i++){
- output[i] = item32_to_fc32(input[i], float(scale_factor));
+ output[i] = item32_sc16_to_fc32(input[i], float(scale_factor));
}
}
@@ -182,7 +182,7 @@ DECLARE_CONVERTER(sc16_item32_be, 1, fc32, 1, PRIORITY_CUSTOM){
//dispatch according to alignment
switch (size_t(output) & 0xf){
case 0x8:
- output[i] = item32_to_fc32(uhd::byteswap(input[i]), float(scale_factor)); i++;
+ output[i] = item32_sc16_to_fc32(uhd::byteswap(input[i]), float(scale_factor)); i++;
case 0x0:
convert_item32_1_to_fc32_1_bswap_guts(_)
break;
@@ -191,6 +191,6 @@ DECLARE_CONVERTER(sc16_item32_be, 1, fc32, 1, PRIORITY_CUSTOM){
//convert remainder
for (; i < nsamps; i++){
- output[i] = item32_to_fc32(uhd::byteswap(input[i]), float(scale_factor));
+ output[i] = item32_sc16_to_fc32(uhd::byteswap(input[i]), float(scale_factor));
}
}
diff --git a/host/lib/convert/convert_fc64_with_sse2.cpp b/host/lib/convert/convert_fc64_with_sse2.cpp
index 2093cf476..a4f2df2e7 100644
--- a/host/lib/convert/convert_fc64_with_sse2.cpp
+++ b/host/lib/convert/convert_fc64_with_sse2.cpp
@@ -64,7 +64,7 @@ DECLARE_CONVERTER(fc64, 1, sc16_item32_le, 1, PRIORITY_CUSTOM){
//convert remainder
for (; i < nsamps; i++){
- output[i] = fc64_to_item32(input[i], scale_factor);
+ output[i] = fc64_to_item32_sc16(input[i], scale_factor);
}
}
@@ -110,7 +110,7 @@ DECLARE_CONVERTER(fc64, 1, sc16_item32_be, 1, PRIORITY_CUSTOM){
//convert remainder
for (; i < nsamps; i++){
- output[i] = uhd::byteswap(fc64_to_item32(input[i], scale_factor));
+ output[i] = uhd::byteswap(fc64_to_item32_sc16(input[i], scale_factor));
}
}
@@ -159,7 +159,7 @@ DECLARE_CONVERTER(sc16_item32_le, 1, fc64, 1, PRIORITY_CUSTOM){
//convert remainder
for (; i < nsamps; i++){
- output[i] = item32_to_fc64(input[i], scale_factor);
+ output[i] = item32_sc16_to_fc64(input[i], scale_factor);
}
}
@@ -207,6 +207,6 @@ DECLARE_CONVERTER(sc16_item32_be, 1, fc64, 1, PRIORITY_CUSTOM){
//convert remainder
for (; i < nsamps; i++){
- output[i] = item32_to_fc64(uhd::byteswap(input[i]), scale_factor);
+ output[i] = item32_sc16_to_fc64(uhd::byteswap(input[i]), scale_factor);
}
}
diff --git a/host/lib/convert/gen_convert_general.py b/host/lib/convert/gen_convert_general.py
index 0cd4155fa..43e1f9967 100644
--- a/host/lib/convert/gen_convert_general.py
+++ b/host/lib/convert/gen_convert_general.py
@@ -34,7 +34,7 @@ DECLARE_CONVERTER($(cpu_type), 1, sc16_item32_$(end), 1, PRIORITY_GENERAL){
item32_t *output = reinterpret_cast<item32_t *>(outputs[0]);
for (size_t i = 0; i < nsamps; i++){
- output[i] = $(to_wire)($(cpu_type)_to_item32(input[i], scale_factor));
+ output[i] = $(to_wire)($(cpu_type)_to_item32_sc16(input[i], scale_factor));
}
}
@@ -43,7 +43,35 @@ DECLARE_CONVERTER(sc16_item32_$(end), 1, $(cpu_type), 1, PRIORITY_GENERAL){
$(cpu_type)_t *output = reinterpret_cast<$(cpu_type)_t *>(outputs[0]);
for (size_t i = 0; i < nsamps; i++){
- output[i] = item32_to_$(cpu_type)($(to_host)(input[i]), scale_factor);
+ output[i] = item32_sc16_to_$(cpu_type)($(to_host)(input[i]), scale_factor);
+ }
+}
+
+DECLARE_CONVERTER(sc8_item32_$(end), 1, $(cpu_type), 1, PRIORITY_GENERAL){
+ const item32_t *input = reinterpret_cast<const item32_t *>(size_t(inputs[0]) & ~0x3);
+ $(cpu_type)_t *output = reinterpret_cast<$(cpu_type)_t *>(outputs[0]);
+ $(cpu_type)_t dummy;
+
+ const bool head_case = ((size_t(inputs[0]) & 0x3) != 0);
+ const bool tail_case = ((nsamps & 0x1) == 0)? head_case : not head_case;
+ const size_t num_pairs = (head_case? nsamps-1 : nsamps)/2;
+ size_t i = 0, j = 0;
+
+ //special head case, probably from a partial recv
+ if (head_case){
+ const item32_t item0 = $(to_host)(input[i++]);
+ item32_sc8_to_$(cpu_type)(item0, dummy, output[j++], scale_factor);
+ }
+
+ for (; i < num_pairs; i++, j+=2){
+ const item32_t item_i = $(to_host)(input[i]);
+ item32_sc8_to_$(cpu_type)(item_i, output[j], output[j+1], scale_factor);
+ }
+
+ //special tail case, finished on an odd number
+ if (tail_case){
+ const item32_t item_i = $(to_host)(input[i]);
+ item32_sc8_to_$(cpu_type)(item_i, output[j], dummy, scale_factor);
}
}
"""
--
cgit v1.2.3