christoph/failnix
1
Fork 0
Code Releases Activity

Add wasm tacle-bench targets

Browse Source
This commit is contained in:
Christoph Urlacher
2026-06-12 20:06:22 +02:00
parent 30daa8a00c
commit 08c2e9c13d
1122 changed files with 520422 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

25
targets/wasm-tacle/kernel/matrix1/CMakeLists.txt Normal file
View File

@ -0,0 +1,25 @@
# ~~~
# SPDX-License-Identifier: MIT
# SPDX-FileCopyrightText: 2026, Friedrich-Alexander-Universität Erlangen-Nürnberg (FAU)
# ~~~
cmake_minimum_required(VERSION 3.20)
project(matrix1)
set(TACLEBENCH_ROOT_PATH "${CMAKE_CURRENT_LIST_DIR}/../../..")
set(REPOSITORY_ROOT_PATH "${CMAKE_CURRENT_LIST_DIR}/../../../..")
set(APP_TARGET_NAME "${CMAKE_PROJECT_NAME}")
if(DEFINED TACLEBENCH_VARIANT AND "${TACLEBENCH_VARIANT}" STREQUAL "inline")
set(APP_SOURCE_FILE_PATH
"generated/modified_sources/inline/matrix1.c")
else()
set(APP_SOURCE_FILE_PATH
"generated/modified_sources/default/matrix1.c")
endif()
include(${REPOSITORY_ROOT_PATH}/cmake/taclebench_wasm.cmake)

63
targets/wasm-tacle/kernel/matrix1/ChangeLog.txt Executable file
View File

@ -0,0 +1,63 @@
File: matrix1.c
Original provenience: DSP-Stone benchmark suite, matrix1_fixed
http://www.ice.rwth-aachen.de/research/tools-projects/entry/detail/dspstone
2016-02-09:
- Added TACLeBench header
- Redefined function pin_down() as returning void since the return value is
insignificant
- Function pin_down() fills matrices A and B with 1s and matrix C with 0s.
First call to it is kept as the initilization part, but the second call at
the end of main function seems unnecessary, hence removed
- Added a new main function that first calls init function then the old main
function sans init
- Redefined matrices A, B and C as global variables so that they are not local
to new main function and can be initialized and processed with diferent
functions
- Annotated matrix1_fixed_main() as the entry point of the analysis
- Removed unnecessary dereference operator from '*p_c++;' (line TODO:XXX) to
remove compiler warning 'value computed not used'
- Removed seemingly unnecessary empty lines
- Moved around all the following so that they are in the given order just after
the header
- macro definitions
- forward declaration of functions
- declaration of global variables
- initialization functions
- main functions
- Applied code formatting according to the following rules
- Lines shall not be wider than 80 characters; whenever possible, appropriate
line breaks shall be inserted to keep lines below 80 characters
- Indentation is done using whitespaces only, no tabs. Code is indented by
two whitespaces
- Two empty lines are put between any two functions
- In non-empty lists or index expressions, opening '(' and '[' are followed by
one whitespace, closing ')' and ']' are preceded by one whitespace
- In comma- or colon-separated argument lists, one whitespace is put after
each comma/colon
- Names of functions and global variables all start with a benchmark-specific
prefix (here: statemate_) followed by lowercase letter
- For pointer types, one whitespace is put before the '*'
- Operators within expressions shall be preceded and followed by one
whitespace
- Code of then- and else-parts of if-then-else statements shall be put in
separate lines, not in the same lines as the if-condition or the keyword
"else"
- Opening braces '{' denoting the beginning of code for some if-else or loop
body shall be put at the end of the same line where the keywords "if",
"else", "for", "while" etc. occur
2016-04-26:
- Renamed benchmark: matrix1_fixed -> matrix1
- Changed prefix in function/variable names (matrix1_fixed -> matrix1)
- Removed macro definition (#define TYPE int) and changed 'TYPE' into 'int'
- Removed macro definition (#define STORAGE_CLASS register) and changed 'STORAGE_CLASS'
into 'register'
- Declared initialisation value as volatile
- Added matrix1_return() function and call to this function from matrix1_main()
2016-10-08
- Added prefix to global matrix names: A -> matrix1_A, etc.
2017-08-03
- Fixed off-by-one error in loop header

BIN
targets/wasm-tacle/kernel/matrix1/generated/default/matrix1.wasm Executable file
View File

Binary file not shown.

356
targets/wasm-tacle/kernel/matrix1/generated/default/matrix1.wat Normal file
View File

@ -0,0 +1,356 @@
(module $matrix1.wasm
(type (;0;) (func (param i32 i32)))
(type (;1;) (func))
(type (;2;) (func (result i32)))
(import "__pragma" "loopbound" (func $__pragma_loopbound (type 0)))
(func $__wasm_apply_data_relocs (type 1))
(func $matrix1_main (type 1)
(local i32 i32 i32 i32 i32)
i32.const 10
i32.const 10
call $__pragma_loopbound
i32.const 1824
local.set 0
i32.const 0
local.set 1
loop ;; label = @1
i32.const 10
i32.const 10
call $__pragma_loopbound
local.get 1
i32.const 40
i32.mul
i32.const 1424
i32.add
local.set 2
i32.const -400
local.set 3
loop ;; label = @2
local.get 0
i32.const 0
i32.store
i32.const 10
i32.const 10
call $__pragma_loopbound
local.get 0
local.get 0
i32.load
local.get 2
i32.load
local.get 3
i32.const 1424
i32.add
i32.load
i32.mul
i32.add
local.tee 4
i32.store
local.get 0
local.get 4
local.get 2
i32.load offset=4
local.get 3
i32.const 1428
i32.add
i32.load
i32.mul
i32.add
local.tee 4
i32.store
local.get 0
local.get 4
local.get 2
i32.load offset=8
local.get 3
i32.const 1432
i32.add
i32.load
i32.mul
i32.add
local.tee 4
i32.store
local.get 0
local.get 4
local.get 2
i32.load offset=12
local.get 3
i32.const 1436
i32.add
i32.load
i32.mul
i32.add
local.tee 4
i32.store
local.get 0
local.get 4
local.get 2
i32.load offset=16
local.get 3
i32.const 1440
i32.add
i32.load
i32.mul
i32.add
local.tee 4
i32.store
local.get 0
local.get 4
local.get 2
i32.load offset=20
local.get 3
i32.const 1444
i32.add
i32.load
i32.mul
i32.add
local.tee 4
i32.store
local.get 0
local.get 4
local.get 2
i32.load offset=24
local.get 3
i32.const 1448
i32.add
i32.load
i32.mul
i32.add
local.tee 4
i32.store
local.get 0
local.get 4
local.get 2
i32.load offset=28
local.get 3
i32.const 1452
i32.add
i32.load
i32.mul
i32.add
local.tee 4
i32.store
local.get 0
local.get 4
local.get 2
i32.load offset=32
local.get 3
i32.const 1456
i32.add
i32.load
i32.mul
i32.add
local.tee 4
i32.store
local.get 0
local.get 4
local.get 2
i32.load offset=36
local.get 3
i32.const 1460
i32.add
i32.load
i32.mul
i32.add
i32.store
local.get 0
i32.const 4
i32.add
local.set 0
local.get 3
i32.const 40
i32.add
local.tee 3
br_if 0 (;@2;)
end
local.get 1
i32.const 1
i32.add
local.tee 1
i32.const 10
i32.ne
br_if 0 (;@1;)
end)
(func $__original_main (type 2) (result i32)
(local i32 i32 i32)
global.get $__stack_pointer
i32.const 16
i32.sub
local.tee 0
global.set $__stack_pointer
local.get 0
i32.const 1
i32.store offset=12
i32.const 100
i32.const 100
call $__pragma_loopbound
i32.const -400
local.set 1
loop ;; label = @1
local.get 1
i32.const 1424
i32.add
local.get 0
i32.load offset=12
i32.store
local.get 1
i32.const 1428
i32.add
local.get 0
i32.load offset=12
i32.store
local.get 1
i32.const 1432
i32.add
local.get 0
i32.load offset=12
i32.store
local.get 1
i32.const 1436
i32.add
local.get 0
i32.load offset=12
i32.store
local.get 1
i32.const 1440
i32.add
local.get 0
i32.load offset=12
i32.store
local.get 1
i32.const 20
i32.add
local.tee 1
br_if 0 (;@1;)
end
i32.const 100
i32.const 100
call $__pragma_loopbound
i32.const -400
local.set 1
loop ;; label = @1
local.get 1
i32.const 1824
i32.add
local.get 0
i32.load offset=12
i32.store
local.get 1
i32.const 1828
i32.add
local.get 0
i32.load offset=12
i32.store
local.get 1
i32.const 1832
i32.add
local.get 0
i32.load offset=12
i32.store
local.get 1
i32.const 1836
i32.add
local.get 0
i32.load offset=12
i32.store
local.get 1
i32.const 1840
i32.add
local.get 0
i32.load offset=12
i32.store
local.get 1
i32.const 20
i32.add
local.tee 1
br_if 0 (;@1;)
end
i32.const 100
i32.const 100
call $__pragma_loopbound
i32.const -400
local.set 1
loop ;; label = @1
local.get 1
i32.const 2240
i32.add
i32.const 0
i32.store
local.get 1
i32.const 2232
i32.add
i64.const 0
i64.store align=4
local.get 1
i32.const 2224
i32.add
i64.const 0
i64.store align=4
local.get 1
i32.const 20
i32.add
local.tee 1
br_if 0 (;@1;)
end
call $matrix1_main
i32.const 100
i32.const 100
call $__pragma_loopbound
i32.const 0
local.set 2
i32.const -400
local.set 1
loop ;; label = @1
local.get 1
i32.const 2240
i32.add
i32.load
local.get 1
i32.const 2236
i32.add
i32.load
local.get 1
i32.const 2232
i32.add
i32.load
local.get 1
i32.const 2228
i32.add
i32.load
local.get 1
i32.const 2224
i32.add
i32.load
local.get 2
i32.add
i32.add
i32.add
i32.add
i32.add
local.set 2
local.get 1
i32.const 20
i32.add
local.tee 1
br_if 0 (;@1;)
end
local.get 0
i32.const 16
i32.add
global.set $__stack_pointer
i32.const -1
i32.const 0
local.get 2
i32.const 1000
i32.ne
select)
(table (;0;) 1 1 funcref)
(memory (;0;) 1)
(global $__stack_pointer (mut i32) (i32.const 6320))
(global (;1;) i32 (i32.const 2224))
(global (;2;) i32 (i32.const 6320))
(export "memory" (memory 0))
(export "__wasm_apply_data_relocs" (func $__wasm_apply_data_relocs))
(export "entrypoint" (func $matrix1_main))
(export "main" (func $__original_main))
(export "__data_end" (global 1))
(export "__heap_base" (global 2)))

168
targets/wasm-tacle/kernel/matrix1/generated/modified_sources/default/matrix1.c Normal file
View File

@ -0,0 +1,168 @@
/*
This program is part of the TACLeBench benchmark suite.
Version V 1.x
Name: matrix1
Author: Juan Martinez Velarde
Function: Generic matrix - multiply benchmarking
This program performs a matrix multiplication of the form C=AB,
where A and B are two dimensional matrices of arbitrary dimension.
The only restriction os that the inner dimension of the arrays must
be greater than 1.
A[ X x Y ] * B[ Y x Z ] = C[ X x Z ]
|a11 a12 .. a1y|
|a21 a22 .. a2y|
matrix A[ X x Y ]= |.. .. .. .. |
|a(x-1)1 a(x-1)2 .. a(x-1)y|
|ax1 ax2 .. axy|
|b11 b12 .. b1z|
|b21 b22 .. b2z|
matrix B[ Y x Z ]= |.. .. .. .. |
|b(y-1)1 b(y-1)2 .. b(y-1)z|
|by1 by2 .. byz|
|c11 c12 .. c1z|
|c21 c22 .. c2z|
matrix C[ X x Z ]= |.. .. .. .. |
|c(x-1)1 c(x-1)2 .. c(x-1)z|
|cx1 cx2 .. cxz|
matrix elements are stored as
A[ X x Y ] = { a11, a12, .. , a1y,
a21, a22, .. , a2y,
...,
ax1, ax2, .. , axy}
B[ Y x Z ] = { b11, b21, .., b(y-1)1, by1, b12, b22, .. , b(y-1)z, byz }
C[ X x Z ] = { c11, c21, .. , c(x-1)1, cx1, c12, c22, .. ,c(x-1)z, cxz }
Source: DSP-Stone
http://www.ice.rwth-aachen.de/research/tools-projects/entry/detail/dspstone
Changes: no major functional changes
License: may be used, modified, and re-distributed freely
*/
/*
Macro definitions
*/
// Wasm loop bounds
__attribute__((import_module("__pragma"), import_name("loopbound"))) extern void
__pragma_loopbound(unsigned int min_bound, unsigned int max_bound);
#define X 10 /* first dimension of array A */
#define Y 10 /* second dimension of array A, first dimension of array B */
#define Z 10 /* second dimension of array B */
/*
Forward declaration of functions
*/
void matrix1_pin_down(int A[], int B[], int C[]);
void matrix1_init(void);
__attribute__((noinline)) __attribute__((export_name("entrypoint"))) void
matrix1_main(void);
__attribute__((noinline)) __attribute__((export_name("main"))) int main(void);
/*
Declaration of global variables
*/
int matrix1_A[X * Y];
int matrix1_B[Y * Z];
int matrix1_C[X * Z];
/*
Initialization functions
*/
void
matrix1_pin_down(int A[], int B[], int C[]) {
int i;
volatile int x = 1;
__pragma_loopbound(100, 100);
for (i = 0; i < X * Y; i++)
A[i] = x;
__pragma_loopbound(100, 100);
for (i = 0; i < Y * Z; i++)
B[i] = x;
__pragma_loopbound(100, 100);
for (i = 0; i < X * Z; i++)
C[i] = 0;
}
void
matrix1_init(void) {
matrix1_pin_down(&matrix1_A[0], &matrix1_B[0], &matrix1_C[0]);
}
/*
Return function
*/
int
matrix1_return(void) {
int i;
int checksum = 0;
__pragma_loopbound(100, 100);
for (i = 0; i < X * Z; i++)
checksum += matrix1_C[i];
return (checksum == 1000 ? 0 : -1);
}
/*
Main functions
*/
__attribute__((noinline)) __attribute__((export_name("entrypoint"))) void
matrix1_main(void) {
register int *p_a = &matrix1_A[0];
register int *p_b = &matrix1_B[0];
register int *p_c = &matrix1_C[0];
register int f, i, k;
__pragma_loopbound(10, 10);
for (k = 0; k < Z; k++) {
p_a = &matrix1_A[0]; /* point to the beginning of array A */
__pragma_loopbound(10, 10);
for (i = 0; i < X; i++) {
p_b = &matrix1_B[k * Y]; /* take next column */
*p_c = 0;
__pragma_loopbound(10, 10);
for (f = 0; f < Y; f++) /* do multiply */
*p_c += *p_a++ * *p_b++;
p_c++;
}
}
}
__attribute__((noinline)) __attribute__((export_name("main"))) int
main(void) {
matrix1_init();
matrix1_main();
return matrix1_return();
}

177
targets/wasm-tacle/kernel/matrix1/generated/modified_sources/inline/matrix1.c Normal file
View File

@ -0,0 +1,177 @@
/*
This program is part of the TACLeBench benchmark suite.
Version V 1.x
Name: matrix1
Author: Juan Martinez Velarde
Function: Generic matrix - multiply benchmarking
This program performs a matrix multiplication of the form C=AB,
where A and B are two dimensional matrices of arbitrary dimension.
The only restriction os that the inner dimension of the arrays must
be greater than 1.
A[ X x Y ] * B[ Y x Z ] = C[ X x Z ]
|a11 a12 .. a1y|
|a21 a22 .. a2y|
matrix A[ X x Y ]= |.. .. .. .. |
|a(x-1)1 a(x-1)2 .. a(x-1)y|
|ax1 ax2 .. axy|
|b11 b12 .. b1z|
|b21 b22 .. b2z|
matrix B[ Y x Z ]= |.. .. .. .. |
|b(y-1)1 b(y-1)2 .. b(y-1)z|
|by1 by2 .. byz|
|c11 c12 .. c1z|
|c21 c22 .. c2z|
matrix C[ X x Z ]= |.. .. .. .. |
|c(x-1)1 c(x-1)2 .. c(x-1)z|
|cx1 cx2 .. cxz|
matrix elements are stored as
A[ X x Y ] = { a11, a12, .. , a1y,
a21, a22, .. , a2y,
...,
ax1, ax2, .. , axy}
B[ Y x Z ] = { b11, b21, .., b(y-1)1, by1, b12, b22, .. , b(y-1)z, byz }
C[ X x Z ] = { c11, c21, .. , c(x-1)1, cx1, c12, c22, .. ,c(x-1)z, cxz }
Source: DSP-Stone
http://www.ice.rwth-aachen.de/research/tools-projects/entry/detail/dspstone
Changes: no major functional changes
License: may be used, modified, and re-distributed freely
*/
/*
Macro definitions
*/
// Wasm loop bounds
__attribute__((import_module("__pragma"), import_name("loopbound"))) extern void
__pragma_loopbound(unsigned int min_bound, unsigned int max_bound);
#define X 10 /* first dimension of array A */
#define Y 10 /* second dimension of array A, first dimension of array B */
#define Z 10 /* second dimension of array B */
/*
Forward declaration of functions
*/
__attribute__((always_inline)) static inline void
matrix1_pin_down(int A[], int B[], int C[]);
__attribute__((always_inline)) static inline void matrix1_init(void);
__attribute__((noinline)) __attribute__((export_name("entrypoint")))
__attribute__((noinline)) __attribute__((export_name("entrypoint"))) void
matrix1_main(void);
__attribute__((noinline)) __attribute__((export_name("main")))
__attribute__((noinline)) __attribute__((export_name("main"))) int
main(void);
/*
Declaration of global variables
*/
int matrix1_A[X * Y];
int matrix1_B[Y * Z];
int matrix1_C[X * Z];
/*
Initialization functions
*/
__attribute__((always_inline)) static inline void
matrix1_pin_down(int A[], int B[], int C[]) {
int i;
volatile int x = 1;
__pragma_loopbound(100, 100);
for (i = 0; i < X * Y; i++)
A[i] = x;
__pragma_loopbound(100, 100);
for (i = 0; i < Y * Z; i++)
B[i] = x;
__pragma_loopbound(100, 100);
for (i = 0; i < X * Z; i++)
C[i] = 0;
}
__attribute__((always_inline)) static inline void
matrix1_init(void) {
matrix1_pin_down(&matrix1_A[0], &matrix1_B[0], &matrix1_C[0]);
}
/*
Return function
*/
__attribute__((always_inline)) static inline int
matrix1_return(void) {
int i;
int checksum = 0;
__pragma_loopbound(100, 100);
for (i = 0; i < X * Z; i++)
checksum += matrix1_C[i];
return (checksum == 1000 ? 0 : -1);
}
/*
Main functions
*/
__attribute__((noinline)) __attribute__((export_name("entrypoint")))
__attribute__((noinline)) __attribute__((export_name("entrypoint"))) void
matrix1_main(void) {
register int *p_a = &matrix1_A[0];
register int *p_b = &matrix1_B[0];
register int *p_c = &matrix1_C[0];
register int f, i, k;
__pragma_loopbound(10, 10);
for (k = 0; k < Z; k++) {
p_a = &matrix1_A[0]; /* point to the beginning of array A */
__pragma_loopbound(10, 10);
for (i = 0; i < X; i++) {
p_b = &matrix1_B[k * Y]; /* take next column */
*p_c = 0;
__pragma_loopbound(10, 10);
for (f = 0; f < Y; f++) /* do multiply */
*p_c += *p_a++ * *p_b++;
p_c++;
}
}
}
__attribute__((noinline)) __attribute__((export_name("main")))
__attribute__((noinline)) __attribute__((export_name("main"))) int
main(void) {
matrix1_init();
matrix1_main();
return matrix1_return();
}

169
targets/wasm-tacle/kernel/matrix1/matrix1.c Executable file
View File

@ -0,0 +1,169 @@
/*
This program is part of the TACLeBench benchmark suite.
Version V 1.x
Name: matrix1
Author: Juan Martinez Velarde
Function: Generic matrix - multiply benchmarking
This program performs a matrix multiplication of the form C=AB,
where A and B are two dimensional matrices of arbitrary dimension.
The only restriction os that the inner dimension of the arrays must
be greater than 1.
A[ X x Y ] * B[ Y x Z ] = C[ X x Z ]
|a11 a12 .. a1y|
|a21 a22 .. a2y|
matrix A[ X x Y ]= |.. .. .. .. |
|a(x-1)1 a(x-1)2 .. a(x-1)y|
|ax1 ax2 .. axy|
|b11 b12 .. b1z|
|b21 b22 .. b2z|
matrix B[ Y x Z ]= |.. .. .. .. |
|b(y-1)1 b(y-1)2 .. b(y-1)z|
|by1 by2 .. byz|
|c11 c12 .. c1z|
|c21 c22 .. c2z|
matrix C[ X x Z ]= |.. .. .. .. |
|c(x-1)1 c(x-1)2 .. c(x-1)z|
|cx1 cx2 .. cxz|
matrix elements are stored as
A[ X x Y ] = { a11, a12, .. , a1y,
a21, a22, .. , a2y,
...,
ax1, ax2, .. , axy}
B[ Y x Z ] = { b11, b21, .., b(y-1)1, by1, b12, b22, .. , b(y-1)z, byz }
C[ X x Z ] = { c11, c21, .. , c(x-1)1, cx1, c12, c22, .. ,c(x-1)z, cxz }
Source: DSP-Stone
http://www.ice.rwth-aachen.de/research/tools-projects/entry/detail/dspstone
Changes: no major functional changes
License: may be used, modified, and re-distributed freely
*/
/*
Macro definitions
*/
#define X 10 /* first dimension of array A */
#define Y 10 /* second dimension of array A, first dimension of array B */
#define Z 10 /* second dimension of array B */
/*
Forward declaration of functions
*/
void matrix1_pin_down( int A[ ], int B[ ], int C[ ] );
void matrix1_init( void );
void matrix1_main( void );
int main( void );
/*
Declaration of global variables
*/
int matrix1_A[ X * Y ];
int matrix1_B[ Y * Z ];
int matrix1_C[ X * Z ];
/*
Initialization functions
*/
void matrix1_pin_down( int A[ ], int B[ ], int C[ ] )
{
int i;
volatile int x = 1;
_Pragma( "loopbound min 100 max 100" )
for ( i = 0 ; i < X * Y; i++ )
A[ i ] = x ;
_Pragma( "loopbound min 100 max 100" )
for ( i = 0 ; i < Y * Z ; i++ )
B[ i ] = x ;
_Pragma( "loopbound min 100 max 100" )
for ( i = 0 ; i < X * Z ; i++ )
C[ i ] = 0 ;
}
void matrix1_init( void )
{
matrix1_pin_down( &matrix1_A[ 0 ], &matrix1_B[ 0 ], &matrix1_C[ 0 ] );
}
/*
Return function
*/
int matrix1_return( void )
{
int i;
int checksum = 0;
_Pragma( "loopbound min 100 max 100" )
for ( i = 0; i < X * Z; i++ )
checksum += matrix1_C[ i ];
return ( checksum == 1000 ? 0 : -1 );
}
/*
Main functions
*/
void _Pragma ( "entrypoint" ) matrix1_main( void )
{
register int *p_a = &matrix1_A[ 0 ];
register int *p_b = &matrix1_B[ 0 ];
register int *p_c = &matrix1_C[ 0 ];
register int f, i, k;
_Pragma( "loopbound min 10 max 10" )
for ( k = 0; k < Z; k++ ) {
p_a = &matrix1_A[ 0 ]; /* point to the beginning of array A */
_Pragma( "loopbound min 10 max 10" )
for ( i = 0; i < X; i++ ) {
p_b = &matrix1_B[ k * Y ]; /* take next column */
*p_c = 0;
_Pragma( "loopbound min 10 max 10" )
for ( f = 0; f < Y; f++ ) /* do multiply */
*p_c += *p_a++ * *p_b++;
p_c++;
}
}
}
int main( void )
{
matrix1_init();
matrix1_main();
return matrix1_return();
}