Merge pull request #914 from xmos/ddr_example

DDR example

examples/app_flash_single_model/Makefile

@@ -1,4 +1,4 @@
-TARGET = XCORE-AI-EXPLORER
+TARGET = XCORE-AI-EXPLORER-800
 APP_NAME =
 
 APP_FLAGS  = -report \

examples/app_single_model_on_one_tile_and_DDR/Makefile

@@ -0,0 +1,43 @@
+TARGET = XCORE-AI-EXPLORER-800
+APP_NAME = app_device
+
+APP_FLAGS = -report \
+-O3 -fcomment-asm -mcmodel=large -fxscope -Wno-xcore-fptrgroup \
+-Wno-unknown-pragmas \
+-Wno-bidirectional-buffered-port \
+-lquadflash \
+-g
+#-Wm,--image-size,8388608,--image-base,8388608
+
+# Comment out the below TFLMC_XCORE_PROFILE definition to disable profiling
+APP_FLAGS += -DTFLMC_XCORE_PROFILE
+
+# Enable the below SHARED_TENSOR_ARENA definition for multiple models 
+# or to manage tensor arena array yourself
+#APP_FLAGS += -DSHARED_TENSOR_ARENA
+
+# *******************************************************
+# Include AI Tools library and headers
+# *******************************************************
+# The following variables are defined for AI Tools library 
+# and should be included when building the app.
+# XMOS_AITOOLSLIB_DEFINITIONS
+# XMOS_AITOOLSLIB_INCLUDES
+# XMOS_AITOOLSLIB_LIBRARIES
+#
+ifeq ($(XMOS_AITOOLSLIB_PATH),)
+$(error Path to XMOS AI Tools library and headers not set correctly!)
+endif
+include ${XMOS_AITOOLSLIB_PATH}/buildfiles/aitoolslib.make
+
+XCC_FLAGS = $(APP_FLAGS) $(XMOS_AITOOLSLIB_DEFINITIONS) $(XMOS_AITOOLSLIB_INCLUDES)
+XCC_CPP_FLAGS = $(APP_FLAGS) -std=c++14 $(XMOS_AITOOLSLIB_DEFINITIONS) $(XMOS_AITOOLSLIB_INCLUDES)
+XCC_MAP_FLAGS = $(APP_FLAGS) $(XMOS_AITOOLSLIB_LIBRARIES)
+
+#=============================================================================
+# The following part of the Makefile includes the common build infrastructure
+# for compiling XMOS applications. You should not need to edit below here.
+
+XMOS_MAKE_PATH ?= ../..
+include $(XMOS_MAKE_PATH)/xcommon/module_xcommon/build/Makefile.common
+

examples/app_single_model_on_one_tile_and_DDR/generate_optimized_cpp_for_xcore.py

@@ -0,0 +1,43 @@
+from xmos_ai_tools import xformer
+
+TFLITE_MODEL_PATH = "mobilenetv2.tflite"
+OPTIMIZED_MODEL_PATH = "src/model.tflite"
+
+OPTIMIZED_MODEL_PATH = "src/model.tflite"
+WEIGHT_PARAMS_PATH = "src/model_weights"
+FLASH_IMAGE_PATH = "src/xcore_flash_binary.out"
+print("Generating app cpp files for model...")
+xformer.convert(
+    TFLITE_MODEL_PATH,
+    OPTIMIZED_MODEL_PATH,
+    {
+        "xcore-thread-count": "5",
+        # set conv err threshold
+        "xcore-conv-err-threshold": "3",
+        # operation splitting to reduce tensor arena size
+        "xcore-op-split-tensor-arena": "True",
+        "xcore-op-split-top-op": "0,7",
+        "xcore-op-split-bottom-op": "6,14",
+        "xcore-op-split-num-splits": "8,4",
+        # write weights as an array to be placed in DDR
+        "xcore-write-weights-as-array" : "True",
+        "xcore-weights-in-external-memory" : "True",
+        # For DDR, we want to ideally reduce loads smaller
+        # than 4000 bytes, as they are slower.
+        # But this would increase memory usage on tile
+        # and so it is a tradeoff
+        "xcore-load-externally-if-larger" : "1500",
+        # move weights to this file
+        "xcore-weights-file" : WEIGHT_PARAMS_PATH,
+    },
+)
+xformer.print_optimization_report()
+
+# # Generate flash image to be flashed using xflash
+# xformer.generate_flash(
+#     output_file=FLASH_IMAGE_PATH,
+#     model_files=[OPTIMIZED_MODEL_PATH],
+#     param_files=[WEIGHT_PARAMS_PATH],
+# )
+
+print("Done!")

examples/app_single_model_on_one_tile_and_DDR/src/XCORE-AI-EXPLORER-800.xn

@@ -0,0 +1,104 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<Network xmlns="http://www.xmos.com"
+         xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
+         xsi:schemaLocation="http://www.xmos.com http://www.xmos.com">
+  <Type>Board</Type>
+  <Name>xcore.ai Explorer Kit</Name>
+
+  <Declarations>
+    <Declaration>tileref tile[2]</Declaration>
+  </Declarations>
+
+  <Packages>
+    <Package id="0" Type="XS3-UnA-1024-FB265">
+      <Nodes>
+        <Node Id="0" InPackageId="0" Type="XS3-L16A-1024" Oscillator="24MHz" SystemFrequency="800MHz" ReferenceFrequency="100MHz"
+              SecondaryPllInputDiv="1" SecondaryPllOutputDiv="3" SecondaryPllFeedbackDiv="83">
+          <Boot>
+            <Source Location="bootFlash"/>
+          </Boot>
+          <Extmem sizeMbit="1024" SourcePll="SecondaryPll" Divider="2">
+            <!-- Attributes for Padctrl and Lpddr XML elements are as per equivalently named 'Node Configuration' registers in datasheet -->
+
+            <Padctrl clk="0x30" cke="0x30" cs_n="0x30" we_n="0x30" cas_n="0x30" ras_n="0x30" addr="0x30" ba="0x30" dq="0x31" dqs="0x31" dm="0x30"/>
+            <!--
+              Attributes all have the same meaning, which is:
+              [6] = Schmitt enable, [5] = Slew, [4:3] = drive strength, [2:1] = pull option, [0] = read enable
+
+              Therefore:
+              0x30: 8mA-drive, fast-slew output
+              0x31: 8mA-drive, fast-slew bidir
+            -->
+
+            <Lpddr emr_opcode="0x20"/>
+            <!--
+              Attributes have various meanings:
+              emr_opcode[7:5] = LPDDR drive strength to xcore.ai
+
+              protocol_engine_conf_0[23:21] = tWR clock count at the Extmem Frequency
+              protocol_engine_conf_0[20:15] = tXSR clock count at the Extmem Frequency
+              protocol_engine_conf_0[14:11] = tRAS clock count at the Extmem Frequency
+              protocol_engine_conf_0[10:0]  = tREFI clock count at the Extmem Frequency
+
+              Therefore:
+              0x20: Half drive strength
+              0x30b: tREFI 7.79us, tRAS 0us, tXSR 0us, tWR 0us
+            -->
+          </Extmem>
+          <Tile Number="0" Reference="tile[0]">
+            <Port Location="XS1_PORT_1B" Name="PORT_SQI_CS"/>
+            <Port Location="XS1_PORT_1C" Name="PORT_SQI_SCLK"/>
+            <Port Location="XS1_PORT_4B" Name="PORT_SQI_SIO"/>
+
+            <Port Location="XS1_PORT_4C" Name="PORT_LEDS"/>
+            <Port Location="XS1_PORT_4D" Name="PORT_BUTTONS"/>
+
+            <Port Location="XS1_PORT_1I"  Name="WIFI_WIRQ"/>
+            <Port Location="XS1_PORT_1J"  Name="WIFI_MOSI"/>
+            <Port Location="XS1_PORT_4E"  Name="WIFI_WUP_RST_N"/>
+            <Port Location="XS1_PORT_4F"  Name="WIFI_CS_N"/>
+            <Port Location="XS1_PORT_1L"  Name="WIFI_CLK"/>
+            <Port Location="XS1_PORT_1M"  Name="WIFI_MISO"/>
+          </Tile>
+          <Tile Number="1" Reference="tile[1]">
+            <!-- Mic related ports -->
+            <Port Location="XS1_PORT_1G" Name="PORT_PDM_CLK"/>
+            <Port Location="XS1_PORT_1F" Name="PORT_PDM_DATA"/>
+
+            <!-- Audio ports -->
+            <Port Location="XS1_PORT_1D" Name="PORT_MCLK_IN"/>
+            <Port Location="XS1_PORT_1C" Name="PORT_I2S_BCLK"/>
+            <Port Location="XS1_PORT_1B" Name="PORT_I2S_LRCLK"/>
+            <Port Location="XS1_PORT_1A" Name="PORT_I2S_DAC_DATA"/>
+            <Port Location="XS1_PORT_1N" Name="PORT_I2S_ADC_DATA"/>
+            <Port Location="XS1_PORT_4A" Name="PORT_CODEC_RST_N"/>
+          </Tile>
+        </Node>
+      </Nodes>
+    </Package>
+  </Packages>
+  <Nodes>
+    <Node Id="2" Type="device:" RoutingId="0x8000">
+      <Service Id="0" Proto="xscope_host_data(chanend c);">
+        <Chanend Identifier="c" end="3"/>
+      </Service>
+    </Node>
+  </Nodes>
+  <Links>
+    <Link Encoding="2wire" Delays="5clk" Flags="XSCOPE">
+      <LinkEndpoint NodeId="0" Link="XL0"/>
+      <LinkEndpoint NodeId="2" Chanend="1"/>
+    </Link>
+  </Links>
+  <ExternalDevices>
+    <Device NodeId="0" Tile="0" Class="SQIFlash" Name="bootFlash" PageSize="256" SectorSize="4096" NumPages="16384">
+      <Attribute Name="PORT_SQI_CS" Value="PORT_SQI_CS"/>
+      <Attribute Name="PORT_SQI_SCLK"   Value="PORT_SQI_SCLK"/>
+      <Attribute Name="PORT_SQI_SIO"  Value="PORT_SQI_SIO"/>
+    </Device>
+  </ExternalDevices>
+  <JTAGChain>
+    <JTAGDevice NodeId="0"/>
+  </JTAGChain>
+
+</Network>

examples/app_single_model_on_one_tile_and_DDR/src/config.xscope

@@ -0,0 +1,21 @@
+<?xml version="1.0" encoding="UTF-8"?>
+
+<!-- ======================================================= -->
+<!-- The 'ioMode' attribute on the xSCOPEconfig              -->
+<!-- element can take the following values:                  -->
+<!--   "none", "basic", "timed"                              -->
+<!--                                                         -->
+<!-- The 'type' attribute on Probe                           -->
+<!-- elements can take the following values:                 -->
+<!--   "STARTSTOP", "CONTINUOUS", "DISCRETE", "STATEMACHINE" -->
+<!--                                                         -->
+<!-- The 'datatype' attribute on Probe                       -->
+<!-- elements can take the following values:                 -->
+<!--   "NONE", "UINT", "INT", "FLOAT"                        -->
+<!-- ======================================================= -->
+
+<xSCOPEconfig ioMode="basic" enabled="true">
+    <!-- For example: -->
+    <!-- <Probe name="Probe Name" type="CONTINUOUS" datatype="UINT" units="Value" enabled="true"/> -->
+    <!-- From the target code, call: xscope_int(PROBE_NAME, value); -->
+</xSCOPEconfig>