GPU: Enforce Uniform buffer alignment to 16bytes

This seems to be a requirement and remove some errors in
renderdoc.
This commit is contained in:
Clément Foucault 2019-05-29 16:57:09 +02:00
parent 77f5210f22
commit 092962cf72
4 changed files with 26 additions and 0 deletions

View File

@ -401,8 +401,15 @@ typedef struct EEVEE_ShadowRender {
int shadow_samples_len;
float shadow_samples_len_inv;
float exponent;
float pad2;
} EEVEE_ShadowRender;
BLI_STATIC_ASSERT_ALIGN(EEVEE_Light, 16)
BLI_STATIC_ASSERT_ALIGN(EEVEE_Shadow, 16)
BLI_STATIC_ASSERT_ALIGN(EEVEE_ShadowCube, 16)
BLI_STATIC_ASSERT_ALIGN(EEVEE_ShadowCascade, 16)
BLI_STATIC_ASSERT_ALIGN(EEVEE_ShadowRender, 16)
/* This is just a really long bitflag with special function to access it. */
#define MAX_LIGHTBITS_FIELDS (MAX_LIGHT / 8)
typedef struct EEVEE_LightBits {
@ -685,8 +692,12 @@ typedef struct EEVEE_CommonUniformBuffer {
int hiz_mip_offset; /* int */
int ray_type; /* int */
float ray_depth; /* float */
float pad_common_ubo;
} EEVEE_CommonUniformBuffer;
BLI_STATIC_ASSERT_ALIGN(EEVEE_CommonUniformBuffer, 16)
/* ray_type (keep in sync with rayType) */
#define EEVEE_RAY_CAMERA 0
#define EEVEE_RAY_SHADOW 1

View File

@ -262,8 +262,11 @@ typedef struct GPUSssKernelData {
float kernel[SSS_SAMPLES][4];
float param[3], max_radius;
int samples;
int pad[3];
} GPUSssKernelData;
BLI_STATIC_ASSERT_ALIGN(GPUSssKernelData, 16)
static void sss_calculate_offsets(GPUSssKernelData *kd, int count, float exponent)
{
float step = 2.0f / (float)(count - 1);

View File

@ -76,6 +76,9 @@ static void gpu_uniformbuffer_initialize(GPUUniformBuffer *ubo, const void *data
GPUUniformBuffer *GPU_uniformbuffer_create(int size, const void *data, char err_out[256])
{
/* Make sure that UBO is padded to size of vec4 */
BLI_assert((size % 16) == 0);
GPUUniformBuffer *ubo = MEM_callocN(sizeof(GPUUniformBufferStatic), "GPUUniformBufferStatic");
ubo->size = size;
ubo->bindpoint = -1;
@ -149,6 +152,9 @@ GPUUniformBuffer *GPU_uniformbuffer_dynamic_create(ListBase *inputs, char err_ou
ubo->buffer.size += gputype * sizeof(float);
}
/* Round up to size of vec4 */
ubo->buffer.size = ((ubo->buffer.size + 15) / 16) * 16;
/* Allocate the data. */
ubo->data = MEM_mallocN(ubo->buffer.size, __func__);

View File

@ -25,6 +25,8 @@
#include "DNA_listBase.h"
#include "DNA_ID.h"
#include "BLI_assert.h"
#ifdef __cplusplus
extern "C" {
#endif
@ -141,6 +143,10 @@ typedef struct LightGridCache {
float visibility_bias, visibility_bleed, visibility_range, _pad5;
} LightGridCache;
/* Theses are used as ubo data. They need to be aligned to size of vec4. */
BLI_STATIC_ASSERT_ALIGN(LightProbeCache, 16)
BLI_STATIC_ASSERT_ALIGN(LightGridCache, 16)
/* ------ Eevee Lightcache ------- */
typedef struct LightCacheTexture {