Fixture 175
float matrix kernel
C · 6 functions · 4 lanes · 15 of 24 function-lanes behave identically
4 of 4 lanes have a function that returns a different result after decompilation: clang-O2 (3/6), gcc-O2 (3/6), clang-O0 (4/6), gcc-O0 (5/6).
A small floating-point numeric kernel over CALLER-OWNED buffers: a dot product at both widths, and the 2x2 matrix operations built on it.
This is the shape the SysV ABI stresses hardest — floating values arriving in xmm0..7 while the pointers arrive in the integer registers, and a loop whose accumulator lives in a register the integer corpus never made the decompiler account for. The buffers also give the differential something to compare besides a return value: matrix2_multiply writes its result through a pointer, so a recovery that computes the right numbers and stores them in the wrong cells still fails.
Every count is validated against a literal bound before it indexes anything, every loop is bounded by that count, and the fixed-size matrix entry points touch exactly four elements of buffers the harness allocates with sixteen.
#include <stdint.h>
/* A small floating-point numeric kernel over CALLER-OWNED buffers: a dot
* product at both widths, and the 2x2 matrix operations built on it.
*
* This is the shape the SysV ABI stresses hardest — floating values arriving in
* xmm0..7 while the pointers arrive in the integer registers, and a loop whose
* accumulator lives in a register the integer corpus never made the decompiler
* account for. The buffers also give the differential something to compare
* besides a return value: `matrix2_multiply` writes its result through a
* pointer, so a recovery that computes the right numbers and stores them in the
* wrong cells still fails.
*
* Every count is validated against a literal bound before it indexes anything,
* every loop is bounded by that count, and the fixed-size matrix entry points
* touch exactly four elements of buffers the harness allocates with sixteen. */
#define FP175_MAX_ELEMENTS 16
#define FP175_MATRIX_CELLS 4
static float fp175_dot_f32(const float *left, const float *right,
int32_t count) {
float total = 0.0f;
int32_t index;
for (index = 0; index < count; ++index) {
total += left[index] * right[index];
}
return total;
}
static int fp175_count_ok(int32_t count) {
return count >= 0 && count <= FP175_MAX_ELEMENTS;
}
__attribute__((noinline)) float dot_product_f32(const float *left,
const float *right,
int32_t count) {
if (left == 0 || right == 0 || !fp175_count_ok(count)) {
return 0.0f;
}
return fp175_dot_f32(left, right, count);
}
/* The binary64 twin. Same source, one width up: the element loads, the
* multiply and the accumulator all change instruction, and a recovery that
* types one from the other disagrees on any inexact input. */
__attribute__((noinline)) double dot_product_f64(const double *left,
const double *right,
int32_t count) {
double total = 0.0;
int32_t index;
if (left == 0 || right == 0 || !fp175_count_ok(count)) {
return 0.0;
}
for (index = 0; index < count; ++index) {
total += left[index] * right[index];
}
return total;
}
/* Squares accumulated one width above the elements, so the total keeps digits a
* binary32 accumulator drops. */
__attribute__((noinline)) double sum_of_squares_f32(const float *series,
int32_t count) {
double total = 0.0;
int32_t index;
if (series == 0 || !fp175_count_ok(count)) {
return 0.0;
}
for (index = 0; index < count; ++index) {
total += (double)series[index] * (double)series[index];
}
return total;
}
/* Row-major 2x2: cells[0] cells[1] / cells[2] cells[3]. */
__attribute__((noinline)) float matrix2_determinant(const float *cells) {
if (cells == 0) {
return 0.0f;
}
return cells[0] * cells[3] - cells[1] * cells[2];
}
/* result = left * right, four dot products of length two. `result` is a
* distinct caller-owned buffer, so no cell is read after it is overwritten. */
__attribute__((noinline)) int32_t matrix2_multiply(const float *left,
const float *right,
float *result) {
if (left == 0 || right == 0 || result == 0) {
return -1;
}
result[0] = left[0] * right[0] + left[1] * right[2];
result[1] = left[0] * right[1] + left[1] * right[3];
result[2] = left[2] * right[0] + left[3] * right[2];
result[3] = left[2] * right[1] + left[3] * right[3];
return FP175_MATRIX_CELLS;
}
/* In-place scaling: the buffer mutation IS the observable, and the return value
* only reports how many elements were touched. */
__attribute__((noinline)) int32_t scale_series_f32(float *series, int32_t count,
float factor) {
int32_t index;
if (series == 0 || !fp175_count_ok(count)) {
return -1;
}
for (index = 0; index < count; ++index) {
series[index] = series[index] * factor;
}
return count;
}Recovered C
Generated by glaurung decompile --style decbench at b47f6b43.
baseline.json records the result after recompiling the C and calling it beside the
original with seeded inputs.
clang -O2
3/6dot_product_f32 fail 57 lines
// glaurung: dot_product_f32 @ 0x1100
float dot_product_f32(const float * arg0, const float * arg1, int32_t arg2) {
int index;
float total;
long ret;
long var10;
float var11;
long var4;
long var53;
float var54;
long var61;
long var62;
long var64;
float var65;
long var8;
ret = 0;
if ((arg0 != 0)) {
if ((arg1 == 0)) {
return ((union { unsigned int bits; float value; }){ .bits = (unsigned int)(ret) }).value;
}
ret = (unsigned long)((unsigned int)((arg2 - 1)));
if (((unsigned long)(15) < (unsigned long)((unsigned long)((unsigned int)(ret))))) {
return ((union { unsigned int bits; float value; }){ .bits = (unsigned int)(ret) }).value;
}
var4 = (unsigned long)((unsigned int)(arg2));
var8 = (unsigned long)((unsigned int)(((unsigned long)((unsigned int)(arg2)) & 3)));
if (((unsigned long)(3) <= (unsigned long)(((unsigned long)((unsigned int)(arg2)) - 1)))) {
var10 = (unsigned long)((unsigned int)((var4 & -4)));
var11 = ((union { unsigned int bits; float value; }){ .bits = (unsigned int)(0) }).value;
index = 0;
do {
ret = ((union { unsigned int bits; float value; }){ .value = ((*(float *)(((long)arg0 + index * 4 + 0xc)) * *(float *)(((long)arg1 + index * 4 + 0xc))) + ((*(float *)(((long)arg0 + index * 4 + 0x8)) * *(float *)(((long)arg1 + index * 4 + 0x8))) + ((*(float *)(((long)arg0 + index * 4 + 0x4)) * *(float *)(((long)arg1 + index * 4 + 0x4))) + ((*(float *)(((long)arg0 + index * 4)) * *(float *)(((long)arg1 + index * 4))) + var11)))) }).bits;
index = (index + 4);
var11 = ((union { unsigned int bits; float value; }){ .bits = (unsigned int)(ret) }).value;
var53 = (unsigned long)((unsigned int)(index));
var54 = ((union { unsigned int bits; float value; }){ .bits = (unsigned int)(ret) }).value;
} while ((var10 != index));
} else {
ret = 0;
var53 = 0;
var54 = ((union { unsigned int bits; float value; }){ .bits = (unsigned int)(0) }).value;
}
if ((var8 == 0)) {
return ((union { unsigned int bits; float value; }){ .bits = (unsigned int)(ret) }).value;
}
var61 = (long)(((long)arg1 + (var53 * 4)));
var62 = (long)(((long)arg0 + (var53 * 4)));
var64 = 0;
var65 = var54;
do {
ret = ((union { unsigned int bits; float value; }){ .value = (var65 + (*(float *)((var62 + var64 * 4)) * *(float *)((var61 + var64 * 4)))) }).bits;
var64 = (var64 + 1);
var65 = ((union { unsigned int bits; float value; }){ .bits = (unsigned int)(ret) }).value;
} while ((var8 != var64));
}
return ((union { unsigned int bits; float value; }){ .bits = (unsigned int)(ret) }).value;
}dot_product_f64 pass 56 lines
// glaurung: dot_product_f64 @ 0x11d0
double dot_product_f64(const double * arg0, const double * arg1, int32_t arg2) {
int index;
double total;
long ret;
long var11;
double var12;
long var5;
long var62;
double var63;
long var70;
long var71;
long var73;
double var74;
long var9;
ret = 0;
if (((unsigned long)(0xfffffff0) <= (unsigned long)((unsigned long)((unsigned int)((arg2 - 17)))))) {
if ((arg0 == 0)) {
return ((union { unsigned long long bits; double value; }){ .bits = (unsigned long long)(ret) }).value;
}
if ((arg1 == 0)) {
return ((union { unsigned long long bits; double value; }){ .bits = (unsigned long long)(ret) }).value;
}
var5 = (unsigned long)((unsigned int)(arg2));
var9 = (unsigned long)((unsigned int)(((unsigned long)((unsigned int)(arg2)) & 3)));
if (((unsigned long)(3) <= (unsigned long)(((unsigned long)((unsigned int)(arg2)) - 1)))) {
var11 = (unsigned long)((unsigned int)((var5 & -4)));
var12 = ((union { unsigned long long bits; double value; }){ .bits = (unsigned long long)(0) }).value;
index = 0;
do {
ret = ((union { unsigned long long bits; double value; }){ .value = ((*(double *)(((long)arg0 + index * 8 + 0x18)) * *(double *)(((long)arg1 + index * 8 + 0x18))) + ((*(double *)(((long)arg0 + index * 8 + 0x10)) * *(double *)(((long)arg1 + index * 8 + 0x10))) + ((*(double *)(((long)arg0 + index * 8 + 0x8)) * *(double *)(((long)arg1 + index * 8 + 0x8))) + ((*(double *)(((long)arg0 + index * 8)) * *(double *)(((long)arg1 + index * 8))) + var12)))) }).bits;
index = (index + 4);
var12 = ((union { unsigned long long bits; double value; }){ .bits = (unsigned long long)(ret) }).value;
var62 = (unsigned long)((unsigned int)(index));
var63 = ((union { unsigned long long bits; double value; }){ .bits = (unsigned long long)(ret) }).value;
} while ((var11 != index));
} else {
ret = 0;
var62 = 0;
var63 = ((union { unsigned long long bits; double value; }){ .bits = (unsigned long long)(0) }).value;
}
if ((var9 == 0)) {
return ((union { unsigned long long bits; double value; }){ .bits = (unsigned long long)(ret) }).value;
}
var70 = (long)(((long)arg1 + (var62 * 8)));
var71 = (long)(((long)arg0 + (var62 * 8)));
var73 = 0;
var74 = var63;
do {
ret = ((union { unsigned long long bits; double value; }){ .value = (var74 + (*(double *)((var71 + var73 * 8)) * *(double *)((var70 + var73 * 8)))) }).bits;
var73 = (var73 + 1);
var74 = ((union { unsigned long long bits; double value; }){ .bits = (unsigned long long)(ret) }).value;
} while ((var9 != var73));
}
return ((union { unsigned long long bits; double value; }){ .bits = (unsigned long long)(ret) }).value;
}matrix2_determinant pass 7 lines
// glaurung: matrix2_determinant @ 0x1360
float matrix2_determinant(const float * arg0) {
if ((arg0 == 0)) {
return ((union { unsigned int bits; float value; }){ .bits = (unsigned int)(0) }).value;
}
return ((*(float *)(((long)arg0)) * *(float *)(((long)arg0 + 0xc))) - (*(float *)(((long)arg0 + 0x8)) * *(float *)(((long)arg0 + 0x4))));
}matrix2_multiply fail 18 lines
// glaurung: matrix2_multiply @ 0x1390
int32_t matrix2_multiply(const float * arg0, const float * arg1, float * arg2) {
long ret;
ret = 0xffffffff;
if ((arg0 != 0)) {
if ((arg1 == 0)) {
return ret;
}
if ((arg2 != 0)) {
*(int *)(((long)arg2)) = ((*(int *)(((long)arg0)) * *(int *)(((long)arg1))) + (*(int *)(((long)arg0 + 0x4)) * *(int *)(((long)arg1 + 0x8))));
*(int *)(((long)arg2 + 0x4)) = ((*(int *)(((long)arg0)) * *(int *)(((long)arg1 + 0x4))) + (*(int *)(((long)arg0 + 0x4)) * *(int *)(((long)arg1 + 0xc))));
*(int *)(((long)arg2 + 0x8)) = ((*(int *)(((long)arg0 + 0x8)) * *(int *)(((long)arg1))) + (*(int *)(((long)arg0 + 0xc)) * *(int *)(((long)arg1 + 0x8))));
*(int *)(((long)arg2 + 0xc)) = ((*(int *)(((long)arg0 + 0x8)) * *(int *)(((long)arg1 + 0x4))) + (*(int *)(((long)arg0 + 0xc)) * *(int *)(((long)arg1 + 0xc))));
ret = 4;
}
}
return ret;
}scale_series_f32 fail 86 lines
// glaurung: scale_series_f32 @ 0x1420
int32_t scale_series_f32(float * arg0, int32_t arg1, float arg2) {
extern long __unknown(long, ...);
int index;
long ret;
long var1;
int var19;
int var20;
int var21;
int var23;
int var24;
int var25;
int var27;
int var28;
int var29;
int var31;
int var32;
int var33;
float var39;
long var4;
ret = 0xffffffff;
if ((arg0 == 0)) {
return ret;
}
if (((unsigned long)(16) < (unsigned long)((unsigned long)((unsigned int)(arg1))))) {
return ret;
}
if (((unsigned long)((unsigned int)(arg1)) == 0)) {
return 0;
}
var1 = (unsigned long)((unsigned int)(arg1));
if (((unsigned long)(4) <= (unsigned long)((unsigned long)((unsigned int)(arg1))))) {
var4 = (unsigned long)((unsigned int)(((unsigned long)((unsigned int)(var1)) & -4)));
var19 = *(int *)(((long)arg0 + 0x4));
var20 = *(int *)(((long)arg0 + 0x8));
var21 = *(int *)(((long)arg0 + 0xc));
/* asm: mulps */
*(int *)(((long)arg0)) = *(int *)(((long)arg0));
*(int *)(((long)arg0 + 0x4)) = var19;
*(int *)(((long)arg0 + 0x8)) = var20;
*(int *)(((long)arg0 + 0xc)) = var21;
if ((var4 != 4)) {
var23 = *(int *)(((long)arg0 + 0x14));
var24 = *(int *)(((long)arg0 + 0x18));
var25 = *(int *)(((long)arg0 + 0x1c));
/* asm: mulps */
*(int *)(((long)arg0 + 0x10)) = *(int *)(((long)arg0 + 0x10));
*(int *)(((long)arg0 + 0x14)) = var23;
*(int *)(((long)arg0 + 0x18)) = var24;
*(int *)(((long)arg0 + 0x1c)) = var25;
if (((unsigned long)((unsigned int)(var4)) != 8)) {
var27 = *(int *)(((long)arg0 + 0x24));
var28 = *(int *)(((long)arg0 + 0x28));
var29 = *(int *)(((long)arg0 + 0x2c));
/* asm: mulps */
*(int *)(((long)arg0 + 0x20)) = *(int *)(((long)arg0 + 0x20));
*(int *)(((long)arg0 + 0x24)) = var27;
*(int *)(((long)arg0 + 0x28)) = var28;
*(int *)(((long)arg0 + 0x2c)) = var29;
if (((unsigned long)((unsigned int)(var4)) != 12)) {
var31 = *(int *)(((long)arg0 + 0x34));
var32 = *(int *)(((long)arg0 + 0x38));
var33 = *(int *)(((long)arg0 + 0x3c));
/* asm: mulps */
*(int *)(((long)arg0 + 0x30)) = *(int *)(((long)arg0 + 0x30));
*(int *)(((long)arg0 + 0x34)) = var31;
*(int *)(((long)arg0 + 0x38)) = var32;
*(int *)(((long)arg0 + 0x3c)) = var33;
}
}
}
if ((var4 != var1)) {
L_1490: ;
do {
var39 = ((union { unsigned int bits; float value; }){ .bits = (unsigned int)(__unknown(0)) }).value;
*(float *)(((long)arg0 + var4 * 4)) = var39;
index = (var4 + 1);
var4 = (unsigned long)((unsigned int)(index));
} while ((var1 != index));
}
} else {
var4 = 0;
goto L_1490;
}
return (unsigned int)(arg1);
}sum_of_squares_f32 pass 61 lines
// glaurung: sum_of_squares_f32 @ 0x12a0
double sum_of_squares_f32(const float * arg0, int32_t arg1) {
int index;
double total;
long ret;
long var11;
double var12;
double var29;
double var43;
long var5;
double var57;
double var71;
long var79;
double var80;
long var87;
long var89;
long var9;
double var90;
double var96;
ret = 0;
if (((unsigned long)((unsigned long)((unsigned int)((arg1 - 1)))) <= (unsigned long)(15))) {
if ((arg0 == 0)) {
return ((union { unsigned long long bits; double value; }){ .bits = (unsigned long long)(ret) }).value;
}
var5 = (unsigned long)((unsigned int)(arg1));
var9 = (unsigned long)((unsigned int)(((unsigned long)((unsigned int)(arg1)) & 3)));
if (((unsigned long)(3) <= (unsigned long)(((unsigned long)((unsigned int)(arg1)) - 1)))) {
var11 = (unsigned long)((unsigned int)((var5 & -4)));
var12 = ((union { unsigned long long bits; double value; }){ .bits = (unsigned long long)(0) }).value;
index = 0;
do {
var29 = (double)(*(float *)(((long)arg0 + index * 4)));
var43 = (double)(*(float *)(((long)arg0 + index * 4 + 0x4)));
var57 = (double)(*(float *)(((long)arg0 + index * 4 + 0x8)));
var71 = (double)(*(float *)(((long)arg0 + index * 4 + 0xc)));
ret = ((union { unsigned long long bits; double value; }){ .value = ((var71 * var71) + ((var57 * var57) + ((var43 * var43) + ((var29 * var29) + var12)))) }).bits;
index = (index + 4);
var12 = ((union { unsigned long long bits; double value; }){ .bits = (unsigned long long)(ret) }).value;
var79 = (unsigned long)((unsigned int)(index));
var80 = ((union { unsigned long long bits; double value; }){ .bits = (unsigned long long)(ret) }).value;
} while ((var11 != index));
} else {
ret = 0;
var79 = 0;
var80 = ((union { unsigned long long bits; double value; }){ .bits = (unsigned long long)(0) }).value;
}
if ((var9 == 0)) {
return ((union { unsigned long long bits; double value; }){ .bits = (unsigned long long)(ret) }).value;
}
var87 = (long)(((long)arg0 + (var79 * 4)));
var89 = 0;
var90 = var80;
do {
var96 = (double)(*(float *)((var87 + var89 * 4)));
ret = ((union { unsigned long long bits; double value; }){ .value = (var90 + (var96 * var96)) }).bits;
var89 = (var89 + 1);
var90 = ((union { unsigned long long bits; double value; }){ .bits = (unsigned long long)(ret) }).value;
} while ((var9 != var89));
}
return ((union { unsigned long long bits; double value; }){ .bits = (unsigned long long)(ret) }).value;
}gcc -O2
3/6dot_product_f32 fail 29 lines
// glaurung: dot_product_f32 @ 0x1100
float dot_product_f32(const float * arg0, const float * arg1, int32_t arg2) {
int index;
float total;
long ret;
long var10;
float var13;
if ((arg0 == 0)) {
return ((union { unsigned int bits; float value; }){ .bits = (unsigned int)(0) }).value;
}
if ((arg1 == 0)) {
return ((union { unsigned int bits; float value; }){ .bits = (unsigned int)(0) }).value;
}
if (((unsigned long)(15) < (unsigned long)((unsigned long)((unsigned int)((arg2 - 1)))))) {
return ((union { unsigned int bits; float value; }){ .bits = (unsigned int)(0) }).value;
}
var10 = (unsigned long)((unsigned int)(arg2));
index = 0;
var13 = ((union { unsigned int bits; float value; }){ .bits = (unsigned int)(0) }).value;
do {
ret = *(int *)(((long)arg0 + index * 4));
ret = (ret * *(int *)(((long)arg1 + index * 4)));
ret = ((unsigned long)((unsigned int)(index)) + 1);
var13 = (var13 + ((union { unsigned int bits; float value; }){ .bits = (unsigned int)(ret) }).value);
index = ret;
} while ((var10 != ret));
ret = ((union { unsigned int bits; float value; }){ .value = var13 }).bits;
return var13;
}dot_product_f64 fail 29 lines
// glaurung: dot_product_f64 @ 0x1150
double dot_product_f64(const double * arg0, const double * arg1, int32_t arg2) {
int index;
double total;
long ret;
long var10;
double var13;
if ((arg0 == 0)) {
return ((union { unsigned long long bits; double value; }){ .bits = (unsigned long long)(0) }).value;
}
if ((arg1 == 0)) {
return ((union { unsigned long long bits; double value; }){ .bits = (unsigned long long)(0) }).value;
}
if (((unsigned long)(15) < (unsigned long)((unsigned long)((unsigned int)((arg2 - 1)))))) {
return ((union { unsigned long long bits; double value; }){ .bits = (unsigned long long)(0) }).value;
}
var10 = (unsigned long)((unsigned int)(arg2));
index = 0;
var13 = ((union { unsigned long long bits; double value; }){ .bits = (unsigned long long)(0) }).value;
do {
ret = *(long *)(((long)arg0 + index * 8));
ret = (ret * *(long *)(((long)arg1 + index * 8)));
ret = ((unsigned long)((unsigned int)(index)) + 1);
var13 = (var13 + ((union { unsigned long long bits; double value; }){ .bits = (unsigned long long)(ret) }).value);
index = ret;
} while ((var10 != ret));
ret = ((union { unsigned long long bits; double value; }){ .value = var13 }).bits;
return var13;
}matrix2_determinant pass 7 lines
// glaurung: matrix2_determinant @ 0x11f0
float matrix2_determinant(const float * arg0) {
if ((arg0 == 0)) {
return ((union { unsigned int bits; float value; }){ .bits = (unsigned int)(0) }).value;
}
return ((*(float *)(((long)arg0)) * *(float *)(((long)arg0 + 0xc))) - (*(float *)(((long)arg0 + 0x4)) * *(float *)(((long)arg0 + 0x8))));
}matrix2_multiply fail 13 lines
// glaurung: matrix2_multiply @ 0x1220
int32_t matrix2_multiply(const float * arg0, const float * arg1, float * arg2) {
if (((arg1 != 0) && (arg2 != 0))) {
if ((arg0 != 0)) {
*(int *)(((long)arg2)) = ((*(int *)(((long)arg0)) * *(int *)(((long)arg1))) + (*(int *)(((long)arg0 + 0x4)) * *(int *)(((long)arg1 + 0x8))));
*(int *)(((long)arg2 + 0x4)) = ((*(int *)(((long)arg0)) * *(int *)(((long)arg1 + 0x4))) + (*(int *)(((long)arg0 + 0x4)) * *(int *)(((long)arg1 + 0xc))));
*(int *)(((long)arg2 + 0x8)) = ((*(int *)(((long)arg0 + 0x8)) * *(int *)(((long)arg1))) + (*(int *)(((long)arg0 + 0xc)) * *(int *)(((long)arg1 + 0x8))));
*(int *)(((long)arg2 + 0xc)) = ((*(int *)(((long)arg0 + 0x8)) * *(int *)(((long)arg1 + 0x4))) + (*(int *)(((long)arg0 + 0xc)) * *(int *)(((long)arg1 + 0xc))));
return 4;
}
}
return 0xffffffff;
}scale_series_f32 pass 25 lines
// glaurung: scale_series_f32 @ 0x12c0
int32_t scale_series_f32(float * arg0, int32_t arg1, float arg2) {
int index;
long ret;
long var1;
long var2;
float var3;
ret = (unsigned long)((unsigned int)(arg1));
if ((arg0 == 0)) {
return 0xffffffff;
}
if (((unsigned long)(16) < (unsigned long)((unsigned long)((unsigned int)(arg1))))) {
return 0xffffffff;
}
if (((unsigned long)((unsigned int)(arg1)) != 0)) {
var1 = (long)((((long)arg0 + ((unsigned long)((unsigned int)((arg1 - 1))) * 4)) + 4));
var2 = (long)arg0;
do {
var3 = *(float *)((var2));
var2 = (var2 + 4);
*(float *)((var2 - 0x4)) = (var3 * arg2);
} while ((var2 != var1));
}
return ret;
}sum_of_squares_f32 pass 26 lines
// glaurung: sum_of_squares_f32 @ 0x11a0
double sum_of_squares_f32(const float * arg0, int32_t arg1) {
int index;
double total;
long var10;
long var11;
double var12;
long var17;
double var23;
if ((arg0 == 0)) {
return ((union { unsigned long long bits; double value; }){ .bits = (unsigned long long)(0) }).value;
}
var10 = (unsigned long)((unsigned int)((arg1 - 1)));
if (((unsigned long)(15) < (unsigned long)((unsigned long)((unsigned int)(var10))))) {
return ((union { unsigned long long bits; double value; }){ .bits = (unsigned long long)(0) }).value;
}
var11 = (long)((((long)arg0 + (var10 * 4)) + 4));
var12 = ((union { unsigned long long bits; double value; }){ .bits = (unsigned long long)(0) }).value;
var17 = (long)arg0;
do {
var17 = (var17 + 4);
var23 = (double)(*(float *)((var17 - 0x4)));
var12 = (var12 + (var23 * var23));
} while ((var17 != var11));
return var12;
}clang -O0
4/6dot_product_f32 pass 29 lines
// glaurung: dot_product_f32 @ 0x1100
float dot_product_f32(const float * arg0, const float * arg1, int32_t arg2) {
extern int fp175_count_ok(int);
extern float fp175_dot_f32(int *, int *, int);
float local_4;
float var11;
int var9;
// x86-64 prologue: save rbp, frame 32 bytes
if ((arg0 == 0)) {
local_4 = ((union { unsigned int bits; float value; }){ .bits = (unsigned int)(0) }).value;
// x86-64 epilogue: restore rbp
return local_4;
}
if ((arg1 == 0)) {
local_4 = ((union { unsigned int bits; float value; }){ .bits = (unsigned int)(0) }).value;
// x86-64 epilogue: restore rbp
return local_4;
}
var9 = fp175_count_ok((unsigned long)((unsigned int)(arg2)));
if (((unsigned long)((unsigned int)(var9)) == 0)) {
local_4 = ((union { unsigned int bits; float value; }){ .bits = (unsigned int)(0) }).value;
// x86-64 epilogue: restore rbp
return local_4;
}
var11 = fp175_dot_f32((int *)(arg0), (int *)(arg1), (unsigned long)((unsigned int)(arg2)));
local_4 = var11;
// x86-64 epilogue: restore rbp
return local_4;
}dot_product_f64 pass 32 lines
// glaurung: dot_product_f64 @ 0x1210
double dot_product_f64(const double * arg0, const double * arg1, int32_t arg2) {
extern int fp175_count_ok(int);
double total;
int index;
double local_8;
int var14;
// x86-64 prologue: save rbp, frame 48 bytes
total = ((union { unsigned long long bits; double value; }){ .bits = (unsigned long long)(0) }).value;
if ((arg0 == 0)) {
local_8 = ((union { unsigned long long bits; double value; }){ .bits = (unsigned long long)(0) }).value;
// x86-64 epilogue: restore rbp
return local_8;
}
if ((arg1 == 0)) {
local_8 = ((union { unsigned long long bits; double value; }){ .bits = (unsigned long long)(0) }).value;
// x86-64 epilogue: restore rbp
return local_8;
}
var14 = fp175_count_ok((unsigned long)((unsigned int)(arg2)));
if (((unsigned long)((unsigned int)(var14)) == 0)) {
local_8 = ((union { unsigned long long bits; double value; }){ .bits = (unsigned long long)(0) }).value;
// x86-64 epilogue: restore rbp
return local_8;
}
for (index = 0; (index < arg2); index++) {
total = ((*(double *)(((long)arg0 + ((long)(index) * 8))) * *(double *)(((long)arg1 + ((long)(index) * 8)))) + total);
}
local_8 = total;
// x86-64 epilogue: restore rbp
return local_8;
}matrix2_determinant fail 9 lines
// glaurung: matrix2_determinant @ 0x1380
float matrix2_determinant(const float * arg0) {
// x86-64 prologue: save rbp
if ((arg0 != 0)) {
return ((*(float *)((long)arg0) * *(float *)(((long)arg0 + 0xc))) + ((union { unsigned int bits; float value; }){ .bits = (unsigned int)((unsigned long)((unsigned int)(((unsigned long)((unsigned int)((*(int *)(((long)arg0 + 0x4)) * *(int *)(((long)arg0 + 0x8))))) ^ -0x80000000LL)))) }).value);
} else {
return ((union { unsigned int bits; float value; }){ .bits = (unsigned int)(0) }).value;
}
}matrix2_multiply fail 26 lines
// glaurung: matrix2_multiply @ 0x13f0
int32_t matrix2_multiply(const float * arg0, const float * arg1, float * arg2) {
int local_4;
// x86-64 prologue: save rbp
if ((arg0 == 0)) {
local_4 = -1;
// x86-64 epilogue: restore rbp
return (unsigned int)(local_4);
}
if ((arg1 == 0)) {
local_4 = -1;
// x86-64 epilogue: restore rbp
return (unsigned int)(local_4);
}
if ((arg2 == 0)) {
local_4 = -1;
// x86-64 epilogue: restore rbp
return (unsigned int)(local_4);
}
*(int *)((long)arg2) = ((*(int *)((long)arg0) * *(int *)((long)arg1)) + (*(int *)(((long)arg0 + 0x4)) * *(int *)(((long)arg1 + 0x8))));
*(int *)(((long)arg2 + 0x4)) = ((*(int *)((long)arg0) * *(int *)(((long)arg1 + 0x4))) + (*(int *)(((long)arg0 + 0x4)) * *(int *)(((long)arg1 + 0xc))));
*(int *)(((long)arg2 + 0x8)) = ((*(int *)(((long)arg0 + 0x8)) * *(int *)((long)arg1)) + (*(int *)(((long)arg0 + 0xc)) * *(int *)(((long)arg1 + 0x8))));
*(int *)(((long)arg2 + 0xc)) = ((*(int *)(((long)arg0 + 0x8)) * *(int *)(((long)arg1 + 0x4))) + (*(int *)(((long)arg0 + 0xc)) * *(int *)(((long)arg1 + 0xc))));
// x86-64 epilogue: restore rbp
return 4;
}scale_series_f32 pass 25 lines
// glaurung: scale_series_f32 @ 0x1510
int32_t scale_series_f32(float * arg0, int32_t arg1, float arg2) {
extern int fp175_count_ok(int);
int index;
int local_4;
int var0;
// x86-64 prologue: save rbp, frame 32 bytes
if ((arg0 == 0)) {
local_4 = -1;
// x86-64 epilogue: restore rbp
return (unsigned int)(local_4);
}
var0 = fp175_count_ok((unsigned long)((unsigned int)(arg1)));
if (((unsigned long)((unsigned int)(var0)) == 0)) {
local_4 = -1;
// x86-64 epilogue: restore rbp
return (unsigned int)(local_4);
}
for (index = 0; (index < arg1); index++) {
arg0[(long)(index)] = (arg0[(long)(index)] * arg2);
}
local_4 = arg1;
// x86-64 epilogue: restore rbp
return (unsigned int)(local_4);
}sum_of_squares_f32 pass 27 lines
// glaurung: sum_of_squares_f32 @ 0x12d0
double sum_of_squares_f32(const float * arg0, int32_t arg1) {
extern int fp175_count_ok(int);
double total;
int index;
double local_8;
int var14;
// x86-64 prologue: save rbp, frame 48 bytes
total = ((union { unsigned long long bits; double value; }){ .bits = (unsigned long long)(0) }).value;
if ((arg0 == 0)) {
local_8 = ((union { unsigned long long bits; double value; }){ .bits = (unsigned long long)(0) }).value;
// x86-64 epilogue: restore rbp
return local_8;
}
var14 = fp175_count_ok((unsigned long)((unsigned int)(arg1)));
if (((unsigned long)((unsigned int)(var14)) == 0)) {
local_8 = ((union { unsigned long long bits; double value; }){ .bits = (unsigned long long)(0) }).value;
// x86-64 epilogue: restore rbp
return local_8;
}
for (index = 0; (index < arg1); index++) {
total = (((double)(*(float *)(((long)arg0 + ((long)(index) * 4)))) * (double)(*(float *)(((long)arg0 + ((long)(index) * 4))))) + total);
}
local_8 = total;
// x86-64 epilogue: restore rbp
return local_8;
}gcc -O0
5/6dot_product_f32 pass 24 lines
// glaurung: dot_product_f32 @ 0x1198
float dot_product_f32(const float * arg0, const float * arg1, int32_t arg2) {
extern int fp175_count_ok(int);
extern float fp175_dot_f32(int *, int *, int);
float var10;
int var6;
// x86-64 prologue: save rbp, frame 24 bytes
if ((arg0 == 0)) {
// x86-64 epilogue: restore rbp
return ((union { unsigned int bits; float value; }){ .bits = (unsigned int)(0) }).value;
}
if ((arg1 == 0)) {
// x86-64 epilogue: restore rbp
return ((union { unsigned int bits; float value; }){ .bits = (unsigned int)(0) }).value;
}
var6 = fp175_count_ok((unsigned long)((unsigned int)(arg2)));
if (((unsigned long)((unsigned int)(var6)) == 0)) {
// x86-64 epilogue: restore rbp
return ((union { unsigned int bits; float value; }){ .bits = (unsigned int)(0) }).value;
}
var10 = fp175_dot_f32((int *)(arg0), (int *)(arg1), (unsigned long)((unsigned int)(arg2)));
// x86-64 epilogue: restore rbp
return ((union { unsigned int bits; float value; }){ .bits = (unsigned int)((unsigned int)(((union { unsigned int bits; float value; }){ .value = var10 }).bits)) }).value;
}dot_product_f64 pass 27 lines
// glaurung: dot_product_f64 @ 0x11f3
double dot_product_f64(const double * arg0, const double * arg1, int32_t arg2) {
extern int fp175_count_ok(int);
double total;
int index;
int var16;
// x86-64 prologue: save rbp, frame 40 bytes
total = ((union { unsigned long long bits; double value; }){ .bits = (unsigned long long)(0) }).value;
if ((arg0 == 0)) {
// x86-64 epilogue: restore rbp
return ((union { unsigned long long bits; double value; }){ .bits = (unsigned long long)(0) }).value;
}
if ((arg1 == 0)) {
// x86-64 epilogue: restore rbp
return ((union { unsigned long long bits; double value; }){ .bits = (unsigned long long)(0) }).value;
}
var16 = fp175_count_ok((unsigned long)((unsigned int)(arg2)));
if (((unsigned long)((unsigned int)(var16)) == 0)) {
// x86-64 epilogue: restore rbp
return ((union { unsigned long long bits; double value; }){ .bits = (unsigned long long)(0) }).value;
}
for (index = 0; (index < arg2); index++) {
total = ((*(double *)(((long)arg1 + ((long)(index) * 8))) * *(double *)(((long)arg0 + ((long)(index) * 8)))) + total);
}
// x86-64 epilogue: restore rbp
return total;
}matrix2_determinant pass 9 lines
// glaurung: matrix2_determinant @ 0x1348
float matrix2_determinant(const float * arg0) {
// x86-64 prologue: save rbp
if ((arg0 != 0)) {
return ((*(float *)((arg0 + 3)) * *(float *)((long)arg0)) - (*(float *)((arg0 + 2)) * *(float *)((arg0 + 1))));
} else {
return ((union { unsigned int bits; float value; }){ .bits = (unsigned int)(0) }).value;
}
}matrix2_multiply fail 27 lines
// glaurung: matrix2_multiply @ 0x139b
int32_t matrix2_multiply(const float * arg0, const float * arg1, float * arg2) {
long ret;
// x86-64 prologue: save rbp
if ((arg0 == 0)) {
// x86-64 epilogue: restore rbp
return 0xffffffff;
}
if ((arg1 == 0)) {
// x86-64 epilogue: restore rbp
return 0xffffffff;
}
if ((arg2 == 0)) {
// x86-64 epilogue: restore rbp
return 0xffffffff;
}
ret = (long)arg2;
*(int *)((long)arg2) = ((*(int *)((arg1 + 2)) * *(int *)((arg0 + 1))) + (*(int *)((long)arg0) * *(int *)((long)arg1)));
ret = ((*(int *)((arg1 + 3)) * *(int *)((arg0 + 1))) + (*(int *)((long)arg0) * *(int *)((arg1 + 1))));
*(int *)((ret)) = ret;
ret = ((*(int *)((arg1 + 2)) * *(int *)((arg0 + 3))) + (*(int *)((arg0 + 2)) * *(int *)((long)arg1)));
*(int *)((ret)) = ret;
ret = ((*(int *)((arg1 + 3)) * *(int *)((arg0 + 3))) + (*(int *)((arg0 + 2)) * *(int *)((arg1 + 1))));
*(int *)((ret)) = ret;
// x86-64 epilogue: restore rbp
return 4;
}scale_series_f32 pass 24 lines
// glaurung: scale_series_f32 @ 0x14e1
int32_t scale_series_f32(float * arg0, int32_t arg1, float arg2) {
extern int fp175_count_ok(int);
int index;
float local_20;
int var1;
// x86-64 prologue: save rbp, frame 32 bytes
local_20 = arg2;
if ((arg0 == 0)) {
// x86-64 epilogue: restore rbp
return 0xffffffff;
}
var1 = fp175_count_ok((unsigned long)((unsigned int)(arg1)));
if (((unsigned long)((unsigned int)(var1)) == 0)) {
// x86-64 epilogue: restore rbp
return 0xffffffff;
}
for (index = 0; (index < arg1); index++) {
arg2 = (*(float *)(((long)arg0 + ((long)(index) * 4))) * local_20);
arg0[(long)(index)] = arg2;
}
// x86-64 epilogue: restore rbp
return (unsigned int)(arg1);
}sum_of_squares_f32 pass 23 lines
// glaurung: sum_of_squares_f32 @ 0x129d
double sum_of_squares_f32(const float * arg0, int32_t arg1) {
extern int fp175_count_ok(int);
double total;
int index;
int var16;
// x86-64 prologue: save rbp, frame 32 bytes
total = ((union { unsigned long long bits; double value; }){ .bits = (unsigned long long)(0) }).value;
if ((arg0 == 0)) {
// x86-64 epilogue: restore rbp
return ((union { unsigned long long bits; double value; }){ .bits = (unsigned long long)(0) }).value;
}
var16 = fp175_count_ok((unsigned long)((unsigned int)(arg1)));
if (((unsigned long)((unsigned int)(var16)) == 0)) {
// x86-64 epilogue: restore rbp
return ((union { unsigned long long bits; double value; }){ .bits = (unsigned long long)(0) }).value;
}
for (index = 0; (index < arg1); index++) {
total = (((double)(*(float *)(((long)arg0 + (0 + ((long)((int)((unsigned long)((unsigned int)(index)))) * 4))))) * (double)(*(float *)(((long)arg0 + (0 + ((long)((int)((unsigned long)((unsigned int)(index)))) * 4)))))) + total);
}
// x86-64 epilogue: restore rbp
return total;
}