Skip to content

Commit

Permalink
Revert "Add debug tag to the GpuArray struct to help identify usage p…
Browse files Browse the repository at this point in the history 
…roblems."

This reverts commit 50e20d5.

This was a misguided attempt at making debugging easier by changing the ABI with a compile definition, which is a really bad idea.  The idea behind this commit may come back at some point in another form.
  • Loading branch information
@abergeron
abergeron committed Nov 13, 2013
1 parent 50e20d5 commit ab5aba2
Show file tree
Hide file tree
Showing 2 changed files with 0 additions and 63 deletions.
3 changes: 0 additions & 3 deletions src/compyte/array.h
View file
Original file line number Diff line number Diff line change
Expand Up @@ -22,9 +22,6 @@ extern "C" {
* Main array structure.
*/
typedef struct _GpuArray {
#ifdef DEBUG
char tag[8];
#endif
/**
* Device data buffer.
*/
Expand Down
60 changes: 0 additions & 60 deletions src/compyte_array.c
View file
Original file line number Diff line number Diff line change
Expand Up @@ -17,23 +17,6 @@
#include "compyte/error.h"
#include "compyte/util.h"

#ifdef DEBUG
#include <assert.h>
#define ARRAY_TAG "gpuarray"

#define TAG_ARRAY(a) memcpy((a)->tag, ARRAY_TAG, 8)
#define ASSERT_ARRAY(a) assert(memcmp((a)->tag, ARRAY_TAG, 8) == 0)
#define ASSERT_NOTARRAY(a) assert(memcmp((a)->tag, ARRAY_TAG, 8) != 0)
#define CLEAR(a) memset((a)->tag, 0, 8);

#else
#define TAG_ARRAY(a)
#define ASSERT_ARRAY(a)
#define ASSERT_NOTARRAY(a)
#define CLEAR(a)

#endif

/*
* Returns the boundaries of an array.
*
Expand Down Expand Up @@ -76,8 +59,6 @@ int GpuArray_empty(GpuArray *a, const compyte_buffer_ops *ops, void *ctx,
unsigned int i;
int res = GA_NO_ERROR;

ASSERT_NOTARRAY(a);

if (ops == NULL)
return GA_INVALID_ERROR;

Expand Down Expand Up @@ -106,7 +87,6 @@ int GpuArray_empty(GpuArray *a, const compyte_buffer_ops *ops, void *ctx,
a->strides = calloc(nd, sizeof(ssize_t));
/* F/C distinction comes later */
a->flags = GA_BEHAVED;
TAG_ARRAY(a);
if (a->dimensions == NULL || a->strides == NULL) {
GpuArray_clear(a);
return GA_MEMORY_ERROR;
Expand Down Expand Up @@ -159,7 +139,6 @@ int GpuArray_fromdata(GpuArray *a, const compyte_buffer_ops *ops,
gpudata *data, size_t offset, int typecode,
unsigned int nd, const size_t *dims,
const ssize_t *strides, int writeable) {
ASSERT_NOTARRAY(a);
if (compyte_get_type(typecode)->typecode != typecode)
return GA_VALUE_ERROR;
a->ops = ops;
Expand All @@ -172,7 +151,6 @@ int GpuArray_fromdata(GpuArray *a, const compyte_buffer_ops *ops,
a->dimensions = calloc(nd, sizeof(size_t));
a->strides = calloc(nd, sizeof(ssize_t));
a->flags = (writeable ? GA_WRITEABLE : 0);
TAG_ARRAY(a);
if (a->dimensions == NULL || a->strides == NULL) {
GpuArray_clear(a);
return GA_MEMORY_ERROR;
Expand Down Expand Up @@ -222,8 +200,6 @@ int GpuArray_copy_from_host(GpuArray *a, const compyte_buffer_ops *ops,
}

int GpuArray_view(GpuArray *v, const GpuArray *a) {
ASSERT_NOTARRAY(v);
ASSERT_ARRAY(a);
v->ops = a->ops;
v->data = a->data;
v->ops->buffer_retain(a->data);
Expand All @@ -233,7 +209,6 @@ int GpuArray_view(GpuArray *v, const GpuArray *a) {
v->flags = a->flags;
v->dimensions = calloc(v->nd, sizeof(size_t));
v->strides = calloc(v->nd, sizeof(ssize_t));
TAG_ARRAY(v);
if (v->dimensions == NULL || v->strides == NULL) {
GpuArray_clear(v);
return GA_MEMORY_ERROR;
Expand All @@ -244,7 +219,6 @@ int GpuArray_view(GpuArray *v, const GpuArray *a) {
}

int GpuArray_sync(GpuArray *a) {
ASSERT_ARRAY(a);
return a->ops->buffer_sync(a->data);
}

Expand All @@ -256,8 +230,6 @@ int GpuArray_index_inplace(GpuArray *a, const ssize_t *starts,
ssize_t *newstrs;
size_t new_offset = a->offset;

ASSERT_ARRAY(a);

if ((starts == NULL) || (stops == NULL) || (steps == NULL))
return GA_VALUE_ERROR;

Expand Down Expand Up @@ -340,9 +312,6 @@ int GpuArray_setarray(GpuArray *a, const GpuArray *v) {
int err = GA_NO_ERROR;
void *t;

ASSERT_ARRAY(a);
ASSERT_ARRAY(v);

if (a->nd < v->nd)
return GA_VALUE_ERROR;

Expand Down Expand Up @@ -434,8 +403,6 @@ int GpuArray_reshape_inplace(GpuArray *a, unsigned int nd,
unsigned int ok;
unsigned int i;

ASSERT_ARRAY(a);

if (ord == GA_ANY_ORDER && GpuArray_ISFORTRAN(a) && a->nd > 1)
ord = GA_F_ORDER;

Expand Down Expand Up @@ -593,8 +560,6 @@ int GpuArray_transpose_inplace(GpuArray *a, const unsigned int *new_axes) {
unsigned int j;
unsigned int k;

ASSERT_ARRAY(a);

newdims = calloc(a->nd, sizeof(size_t));
newstrs = calloc(a->nd, sizeof(ssize_t));
if (newdims == NULL || newstrs == NULL) {
Expand Down Expand Up @@ -635,18 +600,14 @@ int GpuArray_transpose_inplace(GpuArray *a, const unsigned int *new_axes) {
}

void GpuArray_clear(GpuArray *a) {
ASSERT_ARRAY(a);
if (a->data)
a->ops->buffer_release(a->data);
free(a->dimensions);
free(a->strides);
CLEAR(a); // This is redundent, but we might do away with the memset below
memset(a, 0, sizeof(*a));
}

int GpuArray_share(const GpuArray *a, const GpuArray *b) {
ASSERT_ARRAY(a);
ASSERT_ARRAY(b);
if (a->ops != b->ops || a->data != b->data) return 0;
/* XXX: redefine buffer_share to mean: is it possible to share?
and use offset to make sure */
Expand All @@ -655,18 +616,13 @@ int GpuArray_share(const GpuArray *a, const GpuArray *b) {

void *GpuArray_context(const GpuArray *a) {
void *res = NULL;
ASSERT_ARRAY(a);
(void)a->ops->property(NULL, a->data, NULL, GA_BUFFER_PROP_CTX, &res);
return res;
}

int GpuArray_move(GpuArray *dst, const GpuArray *src) {
size_t sz;
unsigned int i;

ASSERT_ARRAY(src);
ASSERT_ARRAY(dst);

if (dst->ops != src->ops)
return GA_INVALID_ERROR;
if (!GpuArray_ISWRITEABLE(dst))
Expand All @@ -689,7 +645,6 @@ int GpuArray_move(GpuArray *dst, const GpuArray *src) {
}

int GpuArray_write(GpuArray *dst, const void *src, size_t src_sz) {
ASSERT_ARRAY(dst);
if (!GpuArray_ISWRITEABLE(dst))
return GA_VALUE_ERROR;
if (!GpuArray_ISONESEGMENT(dst))
Expand All @@ -698,14 +653,12 @@ int GpuArray_write(GpuArray *dst, const void *src, size_t src_sz) {
}

int GpuArray_read(void *dst, size_t dst_sz, const GpuArray *src) {
ASSERT_ARRAY(src);
if (!GpuArray_ISONESEGMENT(src))
return GA_UNSUPPORTED_ERROR;
return src->ops->buffer_read(dst, src->data, src->offset, dst_sz);
}

int GpuArray_memset(GpuArray *a, int data) {
ASSERT_ARRAY(a);
if (!GpuArray_ISONESEGMENT(a))
return GA_UNSUPPORTED_ERROR;
return a->ops->buffer_memset(a->data, a->offset, data);
Expand All @@ -728,8 +681,6 @@ int GpuArray_transfer(GpuArray *res, const GpuArray *a, void *new_ctx,
gpudata *tmp;
int err;

ASSERT_ARRAY(a);

ga_boundaries(&start, &end, a->offset, a->nd, a->dimensions, a->strides);
end += GpuArray_ITEMSIZE(a);

Expand All @@ -746,9 +697,6 @@ int GpuArray_transfer(GpuArray *res, const GpuArray *a, void *new_ctx,
const char *GpuArray_error(const GpuArray *a, int err) {
void *ctx;
int err2 = a->ops->property(NULL, a->data, NULL, GA_BUFFER_PROP_CTX, &ctx);

ASSERT_ARRAY(a);

if (err2 != GA_NO_ERROR) {
/* If CUDA refuses to work after any kind of error in kernels
there is not much we can do about it. */
Expand All @@ -761,8 +709,6 @@ void GpuArray_fprintf(FILE *fd, const GpuArray *a) {
unsigned int i;
int comma = 0;

ASSERT_ARRAY(a);

fprintf(fd, "GpuNdArray <%p, %p, %p> nd=%d\n", a, a->data, a->ops, a->nd);
fprintf(fd, "\tdims: %p, str: %p\n", a->dimensions, a->strides);
fprintf(fd, "\tITEMSIZE: %zd\n", GpuArray_ITEMSIZE(a));
Expand Down Expand Up @@ -793,8 +739,6 @@ void GpuArray_fprintf(FILE *fd, const GpuArray *a) {
int GpuArray_is_c_contiguous(const GpuArray *a) {
size_t size = GpuArray_ITEMSIZE(a);
int i;

ASSERT_ARRAY(a);

for (i = a->nd - 1; i >= 0; i--) {
if (a->strides[i] != size) return 0;
Expand All @@ -808,8 +752,6 @@ int GpuArray_is_f_contiguous(const GpuArray *a) {
size_t size = GpuArray_ITEMSIZE(a);
unsigned int i;

ASSERT_ARRAY(a);

for (i = 0; i < a->nd; i++) {
if (a->strides[i] != size) return 0;
// We suppose that overflow will not happen since data has to fit in memory
Expand All @@ -822,8 +764,6 @@ int GpuArray_is_aligned(const GpuArray *a) {
size_t align = compyte_get_type(a->typecode)->align;
unsigned int i;

ASSERT_ARRAY(a);

if (a->offset % align != 0)
return 0;

Expand Down

0 comments on commit ab5aba2

Please sign in to comment.