From c929ec4ffee197461489c91df8ec1aa827c68854 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Jan=20H=C3=B6nig?= <jan.hoenig@fau.de>
Date: Tue, 9 Nov 2021 19:09:20 +0100
Subject: [PATCH] WIP. TODO get rid of C++. Compile with C99
---
pystencils_benchmark/benchmark.py | 33 ++++++++-
pystencils_benchmark/headers/dummy.c | 2 -
pystencils_benchmark/headers/dummy.s | 45 -------------
pystencils_benchmark/templates/Makefile | 67 +++++++++++++++++++
.../aligned_malloc.h} | 7 +-
.../templates/include_Clang.mk | 24 +++++++
pystencils_benchmark/templates/include_GCC.mk | 24 +++++++
pystencils_benchmark/templates/include_ICC.mk | 14 ++++
pystencils_benchmark/templates/kernel.h | 5 ++
.../templates/kernel_pystencils.c | 18 -----
.../templates/kernel_pystencils.h | 3 -
pystencils_benchmark/templates/main.c | 66 ------------------
.../{headers => templates}/timing.c | 0
.../{headers => templates}/timing.h | 7 +-
test.ipynb | 36 +++-------
15 files changed, 183 insertions(+), 168 deletions(-)
delete mode 100644 pystencils_benchmark/headers/dummy.c
delete mode 100644 pystencils_benchmark/headers/dummy.s
create mode 100644 pystencils_benchmark/templates/Makefile
rename pystencils_benchmark/{headers/kerncraft.h => templates/aligned_malloc.h} (82%)
create mode 100644 pystencils_benchmark/templates/include_Clang.mk
create mode 100644 pystencils_benchmark/templates/include_GCC.mk
create mode 100644 pystencils_benchmark/templates/include_ICC.mk
delete mode 100644 pystencils_benchmark/templates/kernel_pystencils.c
delete mode 100644 pystencils_benchmark/templates/kernel_pystencils.h
rename pystencils_benchmark/{headers => templates}/timing.c (100%)
rename pystencils_benchmark/{headers => templates}/timing.h (79%)
diff --git a/pystencils_benchmark/benchmark.py b/pystencils_benchmark/benchmark.py
index 887913f..f3658df 100644
--- a/pystencils_benchmark/benchmark.py
+++ b/pystencils_benchmark/benchmark.py
@@ -37,6 +37,29 @@ def generate_benchmark(kernel_ast: KernelFunction,
with open(src_path / 'main.c', 'w+') as f:
f.write(kernel_main(kernel_ast))
+ copy_static_files(path)
+
+
+def copy_static_files(path: Path) -> None:
+ src_path = path / 'src'
+ src_path.mkdir(parents=True, exist_ok=True)
+ include_path = path / 'include'
+ include_path.mkdir(parents=True, exist_ok=True)
+
+ env = Environment(loader=PackageLoader('pystencils_benchmark'), undefined=StrictUndefined)
+ files = ['Makefile', 'aligned_malloc.h', 'timing.h', 'timing.c', 'include_Clang.mk', 'include_GCC.mk',
+ 'include_ICC.mk']
+ for file_name in files:
+ template = env.get_template(file_name).render()
+ if file_name[-1] == 'h':
+ target_path = include_path / file_name
+ elif file_name[-1] == 'c':
+ target_path = src_path / file_name
+ else:
+ target_path = path / file_name
+ with open(target_path, 'w+') as f:
+ f.write(template)
+
def kernel_main(kernel: KernelFunction, timing: bool = False):
"""
@@ -86,7 +109,7 @@ def kernel_main(kernel: KernelFunction, timing: bool = False):
fields.append((p.field_name, dtype, elements, size, 0, 0))
call_parameters.append(p.field_name)
- includes = f'#include "{kernel_name}"\n'
+ includes = f'#include "{kernel_name}.h"\n'
jinja_context = {
'kernel_code': generate_c(kernel, dialect=Backend.C),
'kernel_name': kernel_name,
@@ -104,8 +127,10 @@ def kernel_main(kernel: KernelFunction, timing: bool = False):
def kernel_header(kernel_ast: KernelFunction, dialect: Backend = Backend.C) -> str:
function_signature = generate_c(kernel_ast, dialect=dialect, signature_only=True)
+ header_guard = f'_{kernel_ast.function_name.upper()}_H'
jinja_context = {
+ 'header_guard': header_guard,
'function_signature': function_signature,
}
@@ -118,7 +143,11 @@ def kernel_source(kernel_ast: KernelFunction, dialect: Backend = Backend.C) -> s
kernel_name = kernel_ast.function_name
function_source = generate_c(kernel_ast, dialect=dialect)
headers = get_headers(kernel_ast)
- headers.add(f'"{kernel_name}.h"')
+ # TODO remove this after pystencils update
+ if isinstance(headers, list):
+ headers.append(f'"{kernel_name}.h"')
+ else:
+ headers.add(f'"{kernel_name}.h"')
jinja_context = {
'function_source': function_source,
diff --git a/pystencils_benchmark/headers/dummy.c b/pystencils_benchmark/headers/dummy.c
deleted file mode 100644
index a9b0070..0000000
--- a/pystencils_benchmark/headers/dummy.c
+++ /dev/null
@@ -1,2 +0,0 @@
-void dummy(void* a) {}
-int var_false = 0;
diff --git a/pystencils_benchmark/headers/dummy.s b/pystencils_benchmark/headers/dummy.s
deleted file mode 100644
index 084f2f6..0000000
--- a/pystencils_benchmark/headers/dummy.s
+++ /dev/null
@@ -1,45 +0,0 @@
-# mark_description "Intel(R) C Intel(R) 64 Compiler XE for applications running on Intel(R) 64, Version 14.0.0.074 Build 2013071";
-# mark_description "6";
-# mark_description "-std=c99 -S -xHost -O3 -fno-alias -restrict";
- .file "dummy.c"
- .section __TEXT, __text
-L_TXTST0:
-# -- Begin _dummy
-# mark_begin;
- .align 4
- .globl _dummy
-_dummy:
-# parameter 1: %rdi
-L_B1.1: # Preds L_B1.0
-L____tag_value__dummy.1: #1.23
- ret #1.24
- .align 4
-L____tag_value__dummy.3: #
- # LOE
-# mark_end;
- .section __DATA, __data
-# -- End _dummy
- .section __DATA, __data
- .globl _dummy.eh
-// -- Begin SEGMENT __eh_frame
- .section __TEXT,__eh_frame,coalesced,no_toc+strip_static_syms+live_support
-__eh_frame_seg:
-L.__eh_frame_seg:
-EH_frame0:
-L_fde_cie_0:
- .long 0x00000014
- .long 0x00000000
- .long 0x00527a01
- .long 0x01107801
- .long 0x08070c10
- .long 0x01900190
-_dummy.eh:
- .long 0x0000001c
- .long _dummy.eh-L_fde_cie_0+0x4
- .quad L____tag_value__dummy.1-_dummy.eh-0x8
- .set L_Qlab1,L____tag_value__dummy.3-L____tag_value__dummy.1
- .quad L_Qlab1
- .long 0x00000000
- .long 0x00000000
-# End
- .subsections_via_symbols
diff --git a/pystencils_benchmark/templates/Makefile b/pystencils_benchmark/templates/Makefile
new file mode 100644
index 0000000..5490559
--- /dev/null
+++ b/pystencils_benchmark/templates/Makefile
@@ -0,0 +1,67 @@
+TAG = GCC
+
+#CONFIGURE BUILD SYSTEM
+TARGET = benchmark-$(TAG)
+BUILD_DIR = ./$(TAG)
+SRC_DIR = ./src
+MAKE_DIR = ./
+Q ?= @
+
+#DO NOT EDIT BELOW
+include $(MAKE_DIR)/include_$(TAG).mk
+INCLUDES += -I./include
+
+VPATH = $(SRC_DIR)
+ASM = $(patsubst $(SRC_DIR)/%.c, $(BUILD_DIR)/%.s,$(wildcard $(SRC_DIR)/*.c))
+OBJ = $(patsubst $(SRC_DIR)/%.c, $(BUILD_DIR)/%.o,$(wildcard $(SRC_DIR)/*.c))
+CPPFLAGS := $(CPPFLAGS) $(DEFINES) $(INCLUDES)
+
+
+${TARGET}: $(BUILD_DIR) $(OBJ)
+ @echo "===> LINKING $(TARGET)"
+ $(Q)${LINKER} ${LFLAGS} -o $(TARGET) $(OBJ) $(LIBS)
+
+asm: $(BUILD_DIR) $(ASM)
+
+$(BUILD_DIR)/%.o: %.c
+ @echo "===> COMPILE $@"
+ $(Q)$(CC) -c $(CPPFLAGS) $(CFLAGS) $< -o $@
+ $(Q)$(CC) $(CPPFLAGS) -MT $(@:.d=.o) -MM $< > $(BUILD_DIR)/$*.d
+
+$(BUILD_DIR)/%.s: %.c
+ @echo "===> GENERATE ASM $@"
+ $(Q)$(CC) -S $(CPPFLAGS) $(CFLAGS) $< -o $@
+
+tags:
+ @echo "===> GENERATE TAGS"
+ $(Q)ctags -R
+
+
+$(BUILD_DIR):
+ @mkdir $(BUILD_DIR)
+
+ifeq ($(findstring $(MAKECMDGOALS),clean),)
+-include $(OBJ:.o=.d)
+endif
+
+.PHONY: clean distclean debug debug-clean debug-distclean
+
+debug:
+ @make TAG=GCCdebug
+
+debug-clean:
+ @make clean TAG=GCCdebug
+
+debug-distclean:
+ @make distclean TAG=GCCdebug
+
+clean:
+ @echo "===> CLEAN"
+ @rm -rf $(BUILD_DIR)
+ @rm -f tags
+
+distclean: clean
+ @echo "===> DIST CLEAN"
+ @rm -f $(TARGET)
+ @rm -f tags
+
diff --git a/pystencils_benchmark/headers/kerncraft.h b/pystencils_benchmark/templates/aligned_malloc.h
similarity index 82%
rename from pystencils_benchmark/headers/kerncraft.h
rename to pystencils_benchmark/templates/aligned_malloc.h
index 5624537..5353157 100644
--- a/pystencils_benchmark/headers/kerncraft.h
+++ b/pystencils_benchmark/templates/aligned_malloc.h
@@ -1,11 +1,8 @@
-#ifndef _KERNCRAFT_H
-#define _KERNCRAFT_H
+#ifndef _ALIGNED_MALLOC_H
+#define _ALIGNED_MALLOC_H
#include <stddef.h>
#include <stdlib.h>
-#define swap(a, b) {void* t = a; a = b; b = t;}
-
-
inline void* aligned_malloc(size_t size, size_t align) {
// Based on http://stackoverflow.com/q/16376942
void *result;
diff --git a/pystencils_benchmark/templates/include_Clang.mk b/pystencils_benchmark/templates/include_Clang.mk
new file mode 100644
index 0000000..dab3ce3
--- /dev/null
+++ b/pystencils_benchmark/templates/include_Clang.mk
@@ -0,0 +1,24 @@
+CC = clang
+CXX = clang++
+FC = gfortran
+LINKER = $(CXX)
+
+ANSI_CFLAGS = -ansi
+ANSI_CFLAGS += -std=c++0x
+ANSI_CFLAGS += -pedantic
+ANSI_CFLAGS += -Wextra
+
+CFLAGS = -O3 -Wno-format -Wall $(ANSI_CFLAGS)
+# More warning pls
+CFLAGS += -Wfloat-equal -Wundef -Wshadow -Wpointer-arith -Wcast-align -Wstrict-overflow=5 -Wwrite-strings -Waggregate-return
+# Maybe too much warnings
+CFLAGS += -Wcast-qual -Wswitch-default -Wconversion -Wunreachable-code
+CXXFLAGS := $(CFLAGS)
+# Specific C flags
+CFLAGS := $(CFLAGS) -Wstrict-prototypes
+FCFLAGS =
+CPPFLAGS = -std=c++0x
+LFLAGS = #-fopenmp
+DEFINES = -D_GNU_SOURCE -DNDEBUG
+INCLUDES = #-I/usr/lib/gcc/x86_64-linux-gnu/4.8/include/
+LIBS = #-L/usr/lib/x86_64-linux-gnu/libomp.so.5
diff --git a/pystencils_benchmark/templates/include_GCC.mk b/pystencils_benchmark/templates/include_GCC.mk
new file mode 100644
index 0000000..7c0c446
--- /dev/null
+++ b/pystencils_benchmark/templates/include_GCC.mk
@@ -0,0 +1,24 @@
+CC = gcc
+CXX = g++
+FC = gfortran
+LINKER = $(CXX)
+
+ANSI_CFLAGS = -ansi
+ANSI_CFLAGS += -std=c++0x
+ANSI_CFLAGS += -pedantic
+ANSI_CFLAGS += -Wextra
+
+CFLAGS = -O3 -Wno-format -Wall $(ANSI_CFLAGS) -fopenmp
+# More warning pls
+#CFLAGS += -Wfloat-equal -Wundef -Wshadow -Wpointer-arith -Wcast-align -Wstrict-overflow=5 -Wwrite-strings -Waggregate-return
+# Maybe too much warnings
+#CFLAGS += -Wcast-qual -Wswitch-default -Wconversion -Wunreachable-code
+CXXFLAGS := $(CFLAGS)
+# Specific C flags
+CFLAGS := $(CFLAGS) -Wstrict-prototypes
+FCFLAGS =
+CPPFLAGS = -std=c++0x
+LFLAGS = -fopenmp
+DEFINES = -D_GNU_SOURCE -DNDEBUG
+INCLUDES =
+LIBS =
diff --git a/pystencils_benchmark/templates/include_ICC.mk b/pystencils_benchmark/templates/include_ICC.mk
new file mode 100644
index 0000000..0c090fb
--- /dev/null
+++ b/pystencils_benchmark/templates/include_ICC.mk
@@ -0,0 +1,14 @@
+CC = icc
+CXX = icpc
+FC = ifort
+LINKER = $(FC)
+
+CFLAGS = -O3 -std=c99 -xHost
+CXXFLAGS = $(CFLAGS)
+FCFLAGS = -O3 -xHost
+LFLAGS =
+DEFINES = -D_GNU_SOURCE
+INCLUDES =
+LIBS =
+
+
diff --git a/pystencils_benchmark/templates/kernel.h b/pystencils_benchmark/templates/kernel.h
index 634b319..c7826ac 100644
--- a/pystencils_benchmark/templates/kernel.h
+++ b/pystencils_benchmark/templates/kernel.h
@@ -1,4 +1,9 @@
+#ifndef {{header_guard}}
+#define {{header_guard}}
+
#define RESTRICT __restrict__
#define FUNC_PREFIX
{{function_signature}};
+
+#endif
diff --git a/pystencils_benchmark/templates/kernel_pystencils.c b/pystencils_benchmark/templates/kernel_pystencils.c
deleted file mode 100644
index 47fbf7c..0000000
--- a/pystencils_benchmark/templates/kernel_pystencils.c
+++ /dev/null
@@ -1,18 +0,0 @@
-
-#include "kerncraft.h"
-#include <stdlib.h>
-#include <stdint.h>
-#include <stdbool.h>
-#include <math.h>
-#include <stdio.h>
-
-{{ includes }}
-
-#define RESTRICT __restrict__
-#define FUNC_PREFIX
-void dummy(void *);
-void timing(double* wcTime, double* cpuTime);
-extern int var_false;
-
-
-{{kernel_code}}
\ No newline at end of file
diff --git a/pystencils_benchmark/templates/kernel_pystencils.h b/pystencils_benchmark/templates/kernel_pystencils.h
deleted file mode 100644
index 539d51f..0000000
--- a/pystencils_benchmark/templates/kernel_pystencils.h
+++ /dev/null
@@ -1,3 +0,0 @@
-#define FUNC_PREFIX
-
-{{function_signature}}
\ No newline at end of file
diff --git a/pystencils_benchmark/templates/main.c b/pystencils_benchmark/templates/main.c
index 37e95fd..7f7ad47 100644
--- a/pystencils_benchmark/templates/main.c
+++ b/pystencils_benchmark/templates/main.c
@@ -1,4 +1,3 @@
-#include "kerncraft.h"
#include <stdlib.h>
#include <stdint.h>
#include <stdbool.h>
@@ -10,47 +9,7 @@
#define RESTRICT __restrict__
#define FUNC_PREFIX
-void dummy(void *);
void timing(double* wcTime, double* cpuTime);
-extern int var_false;
-
-/* see waLBerla src/field/allocation/AlignedMalloc */
-void *aligned_malloc_with_offset( uint64_t size, uint64_t alignment, uint64_t offset )
-{
- // With 0 alignment this function makes no sense
- // use normal malloc instead
- assert( alignment > 0 );
- // Tests if alignment is power of two (assuming alignment>0)
- assert( !(alignment & (alignment - 1)) );
- assert( offset < alignment );
-
- void *pa; // pointer to allocated memory
- void *ptr; // pointer to usable aligned memory
-
- pa=std::malloc( (size+2*alignment-1 )+sizeof(void *));
- if(!pa)
- return nullptr;
-
- // Find next aligned position, starting at pa+sizeof(void*)-1
- ptr=(void*)( ((size_t)pa+sizeof(void *)+alignment-1) & ~(alignment-1));
- ptr=(void*) ( (char*)(ptr) + alignment - offset);
-
- // Store pointer to real allocated chunk just before usable chunk
- *((void **)ptr-1)=pa;
-
- assert( ((size_t)ptr+offset) % alignment == 0 );
-
- return ptr;
-}
-
-void aligned_free( void *ptr )
-{
- // assume that pointer to real allocated chunk is stored just before
- // chunk that was given to user
- if(ptr)
- std::free(*((void **)ptr-1));
-}
-
int main(int argc, char **argv)
{
@@ -63,24 +22,12 @@ int main(int argc, char **argv)
{%- endif %}
for (unsigned long long i = 0; i < {{elements}}; ++i)
{{field_name}}[i] = 0.23;
-
- if(var_false)
- {
- dummy({{field_name}});
- }
-
{%- endfor %}
-
-
{%- for constantName, dataType in constants %}
-
// Constant {{constantName}}
{{dataType}} {{constantName}};
{{constantName}} = 0.23;
- if(var_false)
- dummy(& {{constantName}});
-
{%- endfor %}
for(int warmup = 1; warmup >= 0; --warmup) {
@@ -97,14 +44,6 @@ int main(int argc, char **argv)
for (; repeat > 0; --repeat)
{
{{kernel_name}}({{call_argument_list}});
-
- // Dummy calls
- {%- for field_name, dataType, elements, size, offset, alignment in fields %}
- if(var_false) dummy((void*){{field_name}});
- {%- endfor %}
- {%- for constantName, dataType in constants %}
- if(var_false) dummy((void*)&{{constantName}});
- {%- endfor %}
}
{%- if timing %}
timing(&wcEndTime, &cpuEndTime);
@@ -115,11 +54,6 @@ int main(int argc, char **argv)
}
{%- for field_name, dataType, elements, size, offset, alignment in fields %}
- {%- if alignment > 0 %}
- aligned_free({{field_name}});
- {%- else %}
delete[] {{field_name}};
- {%- endif %}
-
{%- endfor %}
}
diff --git a/pystencils_benchmark/headers/timing.c b/pystencils_benchmark/templates/timing.c
similarity index 100%
rename from pystencils_benchmark/headers/timing.c
rename to pystencils_benchmark/templates/timing.c
diff --git a/pystencils_benchmark/headers/timing.h b/pystencils_benchmark/templates/timing.h
similarity index 79%
rename from pystencils_benchmark/headers/timing.h
rename to pystencils_benchmark/templates/timing.h
index b063384..6c6ff44 100644
--- a/pystencils_benchmark/headers/timing.h
+++ b/pystencils_benchmark/templates/timing.h
@@ -1,9 +1,12 @@
+#ifndef _TIMING_H
+#define _TIMING_H
+
#include <sys/time.h>
#include <sys/types.h>
#include <stdlib.h>
#include <stdio.h>
#include <sys/resource.h>
-
-
void timing(double* wcTime, double* cpuTime);
+
+#endif
\ No newline at end of file
diff --git a/test.ipynb b/test.ipynb
index d1459d8..e6ac951 100644
--- a/test.ipynb
+++ b/test.ipynb
@@ -2,7 +2,7 @@
"cells": [
{
"cell_type": "code",
- "execution_count": 9,
+ "execution_count": 1,
"metadata": {
"collapsed": true
},
@@ -10,16 +10,16 @@
"source": [
"import numpy as np\n",
"import pystencils as ps\n",
- "from pystencils_benchmark import kernel_header, kernel_source, generate_benchmark\n",
+ "from pystencils_benchmark import generate_benchmark\n",
"from pathlib import Path"
]
},
{
"cell_type": "code",
- "execution_count": 10,
+ "execution_count": 2,
"outputs": [],
"source": [
- "config = ps.CreateKernelConfig()"
+ "config = ps.CreateKernelConfig(function_name='vadd')"
],
"metadata": {
"collapsed": false,
@@ -30,7 +30,7 @@
},
{
"cell_type": "code",
- "execution_count": 11,
+ "execution_count": 3,
"outputs": [],
"source": [
"a, b, c = ps.fields(a=np.ones(4000000), b=np.ones(4000000), c=np.ones(4000000))"
@@ -44,7 +44,7 @@
},
{
"cell_type": "code",
- "execution_count": 12,
+ "execution_count": 4,
"outputs": [],
"source": [
"@ps.kernel\n",
@@ -60,7 +60,7 @@
},
{
"cell_type": "code",
- "execution_count": 13,
+ "execution_count": 5,
"outputs": [
{
"data": {
@@ -72,8 +72,8 @@
},
{
"data": {
- "text/plain": "FUNC_PREFIX void kernel(double * RESTRICT _data_a, double * RESTRICT const _data_b, double * RESTRICT const _data_c)\n{\n for (int64_t ctr_0 = 0; ctr_0 < 4000000; ctr_0 += 1)\n {\n _data_a[ctr_0] = _data_b[ctr_0] + _data_c[ctr_0];\n }\n}",
- "text/html": "<div class=\"highlight\"><pre><span></span><span class=\"n\">FUNC_PREFIX</span><span class=\"w\"> </span><span class=\"kt\">void</span><span class=\"w\"> </span><span class=\"n\">kernel</span><span class=\"p\">(</span><span class=\"kt\">double</span><span class=\"w\"> </span><span class=\"o\">*</span><span class=\"w\"> </span><span class=\"n\">RESTRICT</span><span class=\"w\"> </span><span class=\"n\">_data_a</span><span class=\"p\">,</span><span class=\"w\"> </span><span class=\"kt\">double</span><span class=\"w\"> </span><span class=\"o\">*</span><span class=\"w\"> </span><span class=\"n\">RESTRICT</span><span class=\"w\"> </span><span class=\"k\">const</span><span class=\"w\"> </span><span class=\"n\">_data_b</span><span class=\"p\">,</span><span class=\"w\"> </span><span class=\"kt\">double</span><span class=\"w\"> </span><span class=\"o\">*</span><span class=\"w\"> </span><span class=\"n\">RESTRICT</span><span class=\"w\"> </span><span class=\"k\">const</span><span class=\"w\"> </span><span class=\"n\">_data_c</span><span class=\"p\">)</span><span class=\"w\"></span>\n<span class=\"p\">{</span><span class=\"w\"></span>\n<span class=\"w\"> </span><span class=\"k\">for</span><span class=\"w\"> </span><span class=\"p\">(</span><span class=\"kt\">int64_t</span><span class=\"w\"> </span><span class=\"n\">ctr_0</span><span class=\"w\"> </span><span class=\"o\">=</span><span class=\"w\"> </span><span class=\"mi\">0</span><span class=\"p\">;</span><span class=\"w\"> </span><span class=\"n\">ctr_0</span><span class=\"w\"> </span><span class=\"o\"><</span><span class=\"w\"> </span><span class=\"mi\">4000000</span><span class=\"p\">;</span><span class=\"w\"> </span><span class=\"n\">ctr_0</span><span class=\"w\"> </span><span class=\"o\">+=</span><span class=\"w\"> </span><span class=\"mi\">1</span><span class=\"p\">)</span><span class=\"w\"></span>\n<span class=\"w\"> </span><span class=\"p\">{</span><span class=\"w\"></span>\n<span class=\"w\"> </span><span class=\"n\">_data_a</span><span class=\"p\">[</span><span class=\"n\">ctr_0</span><span class=\"p\">]</span><span class=\"w\"> </span><span class=\"o\">=</span><span class=\"w\"> </span><span class=\"n\">_data_b</span><span class=\"p\">[</span><span class=\"n\">ctr_0</span><span class=\"p\">]</span><span class=\"w\"> </span><span class=\"o\">+</span><span class=\"w\"> </span><span class=\"n\">_data_c</span><span class=\"p\">[</span><span class=\"n\">ctr_0</span><span class=\"p\">];</span><span class=\"w\"></span>\n<span class=\"w\"> </span><span class=\"p\">}</span><span class=\"w\"></span>\n<span class=\"p\">}</span><span class=\"w\"></span>\n</pre></div>\n"
+ "text/plain": "FUNC_PREFIX void vadd(double * RESTRICT _data_a, double * RESTRICT const _data_b, double * RESTRICT const _data_c)\n{\n for (int64_t ctr_0 = 0; ctr_0 < 4000000; ctr_0 += 1)\n {\n _data_a[ctr_0] = _data_b[ctr_0] + _data_c[ctr_0];\n }\n}",
+ "text/html": "<div class=\"highlight\"><pre><span></span><span class=\"n\">FUNC_PREFIX</span><span class=\"w\"> </span><span class=\"kt\">void</span><span class=\"w\"> </span><span class=\"n\">vadd</span><span class=\"p\">(</span><span class=\"kt\">double</span><span class=\"w\"> </span><span class=\"o\">*</span><span class=\"w\"> </span><span class=\"n\">RESTRICT</span><span class=\"w\"> </span><span class=\"n\">_data_a</span><span class=\"p\">,</span><span class=\"w\"> </span><span class=\"kt\">double</span><span class=\"w\"> </span><span class=\"o\">*</span><span class=\"w\"> </span><span class=\"n\">RESTRICT</span><span class=\"w\"> </span><span class=\"k\">const</span><span class=\"w\"> </span><span class=\"n\">_data_b</span><span class=\"p\">,</span><span class=\"w\"> </span><span class=\"kt\">double</span><span class=\"w\"> </span><span class=\"o\">*</span><span class=\"w\"> </span><span class=\"n\">RESTRICT</span><span class=\"w\"> </span><span class=\"k\">const</span><span class=\"w\"> </span><span class=\"n\">_data_c</span><span class=\"p\">)</span><span class=\"w\"></span>\n<span class=\"p\">{</span><span class=\"w\"></span>\n<span class=\"w\"> </span><span class=\"k\">for</span><span class=\"w\"> </span><span class=\"p\">(</span><span class=\"kt\">int64_t</span><span class=\"w\"> </span><span class=\"n\">ctr_0</span><span class=\"w\"> </span><span class=\"o\">=</span><span class=\"w\"> </span><span class=\"mi\">0</span><span class=\"p\">;</span><span class=\"w\"> </span><span class=\"n\">ctr_0</span><span class=\"w\"> </span><span class=\"o\"><</span><span class=\"w\"> </span><span class=\"mi\">4000000</span><span class=\"p\">;</span><span class=\"w\"> </span><span class=\"n\">ctr_0</span><span class=\"w\"> </span><span class=\"o\">+=</span><span class=\"w\"> </span><span class=\"mi\">1</span><span class=\"p\">)</span><span class=\"w\"></span>\n<span class=\"w\"> </span><span class=\"p\">{</span><span class=\"w\"></span>\n<span class=\"w\"> </span><span class=\"n\">_data_a</span><span class=\"p\">[</span><span class=\"n\">ctr_0</span><span class=\"p\">]</span><span class=\"w\"> </span><span class=\"o\">=</span><span class=\"w\"> </span><span class=\"n\">_data_b</span><span class=\"p\">[</span><span class=\"n\">ctr_0</span><span class=\"p\">]</span><span class=\"w\"> </span><span class=\"o\">+</span><span class=\"w\"> </span><span class=\"n\">_data_c</span><span class=\"p\">[</span><span class=\"n\">ctr_0</span><span class=\"p\">];</span><span class=\"w\"></span>\n<span class=\"w\"> </span><span class=\"p\">}</span><span class=\"w\"></span>\n<span class=\"p\">}</span><span class=\"w\"></span>\n</pre></div>\n"
},
"metadata": {},
"output_type": "display_data"
@@ -92,22 +92,8 @@
},
{
"cell_type": "code",
- "execution_count": 14,
- "outputs": [
- {
- "ename": "AttributeError",
- "evalue": "'list' object has no attribute 'add'",
- "output_type": "error",
- "traceback": [
- "\u001B[0;31m---------------------------------------------------------------------------\u001B[0m",
- "\u001B[0;31mAttributeError\u001B[0m Traceback (most recent call last)",
- "\u001B[0;32m/tmp/ipykernel_28091/2748650420.py\u001B[0m in \u001B[0;36m<module>\u001B[0;34m\u001B[0m\n\u001B[1;32m 1\u001B[0m \u001B[0mexample_path\u001B[0m \u001B[0;34m=\u001B[0m \u001B[0mPath\u001B[0m\u001B[0;34m.\u001B[0m\u001B[0mcwd\u001B[0m\u001B[0;34m(\u001B[0m\u001B[0;34m)\u001B[0m \u001B[0;34m/\u001B[0m \u001B[0;34m'example'\u001B[0m\u001B[0;34m\u001B[0m\u001B[0;34m\u001B[0m\u001B[0m\n\u001B[0;32m----> 2\u001B[0;31m \u001B[0mgenerate_benchmark\u001B[0m\u001B[0;34m(\u001B[0m\u001B[0mkernel_vadd\u001B[0m\u001B[0;34m,\u001B[0m \u001B[0mexample_path\u001B[0m\u001B[0;34m)\u001B[0m\u001B[0;34m\u001B[0m\u001B[0;34m\u001B[0m\u001B[0m\n\u001B[0m",
- "\u001B[0;32m~/git/pystencils-benchmark/pystencils_benchmark/benchmark.py\u001B[0m in \u001B[0;36mgenerate_benchmark\u001B[0;34m(kernel_ast, path, dialect)\u001B[0m\n\u001B[1;32m 31\u001B[0m \u001B[0mf\u001B[0m\u001B[0;34m.\u001B[0m\u001B[0mwrite\u001B[0m\u001B[0;34m(\u001B[0m\u001B[0mheader\u001B[0m\u001B[0;34m)\u001B[0m\u001B[0;34m\u001B[0m\u001B[0;34m\u001B[0m\u001B[0m\n\u001B[1;32m 32\u001B[0m \u001B[0;34m\u001B[0m\u001B[0m\n\u001B[0;32m---> 33\u001B[0;31m \u001B[0msource\u001B[0m \u001B[0;34m=\u001B[0m \u001B[0mkernel_source\u001B[0m\u001B[0;34m(\u001B[0m\u001B[0mkernel_ast\u001B[0m\u001B[0;34m,\u001B[0m \u001B[0mdialect\u001B[0m\u001B[0;34m)\u001B[0m\u001B[0;34m\u001B[0m\u001B[0;34m\u001B[0m\u001B[0m\n\u001B[0m\u001B[1;32m 34\u001B[0m \u001B[0;32mwith\u001B[0m \u001B[0mopen\u001B[0m\u001B[0;34m(\u001B[0m\u001B[0msrc_path\u001B[0m \u001B[0;34m/\u001B[0m \u001B[0;34mf'{kernel_name}.c'\u001B[0m\u001B[0;34m,\u001B[0m \u001B[0;34m'w+'\u001B[0m\u001B[0;34m)\u001B[0m \u001B[0;32mas\u001B[0m \u001B[0mf\u001B[0m\u001B[0;34m:\u001B[0m\u001B[0;34m\u001B[0m\u001B[0;34m\u001B[0m\u001B[0m\n\u001B[1;32m 35\u001B[0m \u001B[0mf\u001B[0m\u001B[0;34m.\u001B[0m\u001B[0mwrite\u001B[0m\u001B[0;34m(\u001B[0m\u001B[0msource\u001B[0m\u001B[0;34m)\u001B[0m\u001B[0;34m\u001B[0m\u001B[0;34m\u001B[0m\u001B[0m\n",
- "\u001B[0;32m~/git/pystencils-benchmark/pystencils_benchmark/benchmark.py\u001B[0m in \u001B[0;36mkernel_source\u001B[0;34m(kernel_ast, dialect)\u001B[0m\n\u001B[1;32m 119\u001B[0m \u001B[0mfunction_source\u001B[0m \u001B[0;34m=\u001B[0m \u001B[0mgenerate_c\u001B[0m\u001B[0;34m(\u001B[0m\u001B[0mkernel_ast\u001B[0m\u001B[0;34m,\u001B[0m \u001B[0mdialect\u001B[0m\u001B[0;34m=\u001B[0m\u001B[0mdialect\u001B[0m\u001B[0;34m)\u001B[0m\u001B[0;34m\u001B[0m\u001B[0;34m\u001B[0m\u001B[0m\n\u001B[1;32m 120\u001B[0m \u001B[0mheaders\u001B[0m \u001B[0;34m=\u001B[0m \u001B[0mget_headers\u001B[0m\u001B[0;34m(\u001B[0m\u001B[0mkernel_ast\u001B[0m\u001B[0;34m)\u001B[0m\u001B[0;34m\u001B[0m\u001B[0;34m\u001B[0m\u001B[0m\n\u001B[0;32m--> 121\u001B[0;31m \u001B[0mheaders\u001B[0m\u001B[0;34m.\u001B[0m\u001B[0madd\u001B[0m\u001B[0;34m(\u001B[0m\u001B[0;34mf'\"{kernel_name}.h\"'\u001B[0m\u001B[0;34m)\u001B[0m\u001B[0;34m\u001B[0m\u001B[0;34m\u001B[0m\u001B[0m\n\u001B[0m\u001B[1;32m 122\u001B[0m \u001B[0;34m\u001B[0m\u001B[0m\n\u001B[1;32m 123\u001B[0m jinja_context = {\n",
- "\u001B[0;31mAttributeError\u001B[0m: 'list' object has no attribute 'add'"
- ]
- }
- ],
+ "execution_count": 6,
+ "outputs": [],
"source": [
"example_path = Path.cwd() / 'example'\n",
"generate_benchmark(kernel_vadd, example_path)"
--
GitLab