void decode_and_reconstruct_block_inter (uint8_t *rec, int stride, int size, int qp, uint8_t *pblock,
int16_t *coeffq,int tb_split, qmtx_t ** iwmatrix){
int16_t *rcoeff = thor_alloc(2*MAX_TR_SIZE*MAX_TR_SIZE, 16);
int16_t *rblock = thor_alloc(2*MAX_TR_SIZE*MAX_TR_SIZE, 16);
int16_t *rblock2 = thor_alloc(2*MAX_TR_SIZE*MAX_TR_SIZE, 16);
if (tb_split){
int size2 = size/2;
int i,j,k,index;
for (i=0;i<size;i+=size2){
for (j=0;j<size;j+=size2){
index = 2*(i/size2) + (j/size2);
dequantize (coeffq+index*size2*size2, rcoeff, qp, size2, iwmatrix ? iwmatrix[log2i(size2/4)] : NULL, MAX_QUANT_SIZE);
inverse_transform (rcoeff, rblock2, size2);
/* Copy from compact block of quarter size to full size */
for (k=0;k<size2;k++){
memcpy(rblock+(i+k)*size+j,rblock2+k*size2,size2*sizeof(int16_t));
}
}
}
}
else{
dequantize (coeffq, rcoeff, qp, size, iwmatrix ? iwmatrix[log2i(size/4)] : NULL, MAX_QUANT_SIZE);
inverse_transform (rcoeff, rblock, size);
}
reconstruct_block(rblock,pblock,rec,size,stride);
thor_free(rcoeff);
thor_free(rblock);
thor_free(rblock2);
}
void
nv04_emit_tex_obj(struct gl_context *ctx, int emit)
{
struct nv04_context *nv04 = to_nv04_context(ctx);
const int i = emit - NOUVEAU_STATE_TEX_OBJ0;
struct nouveau_surface *s;
uint32_t format = 0xa0, filter = 0x1010;
if (ctx->Texture.Unit[i]._ReallyEnabled) {
struct gl_texture_object *t = ctx->Texture.Unit[i]._Current;
struct gl_texture_image *ti = t->Image[0][t->BaseLevel];
int lod_max = 1, lod_bias = 0;
if (!nouveau_texture_validate(ctx, t))
return;
s = &to_nouveau_texture(t)->surfaces[t->BaseLevel];
if (t->Sampler.MinFilter != GL_NEAREST &&
t->Sampler.MinFilter != GL_LINEAR) {
lod_max = CLAMP(MIN2(t->Sampler.MaxLod, t->_MaxLambda),
0, 15) + 1;
lod_bias = CLAMP(ctx->Texture.Unit[i].LodBias +
t->Sampler.LodBias, -16, 15) * 8;
}
format |= nvgl_wrap_mode(t->Sampler.WrapT) << 28 |
nvgl_wrap_mode(t->Sampler.WrapS) << 24 |
ti->HeightLog2 << 20 |
ti->WidthLog2 << 16 |
lod_max << 12 |
get_tex_format(ti);
filter |= log2i(t->Sampler.MaxAnisotropy) << 31 |
nvgl_filter_mode(t->Sampler.MagFilter) << 28 |
log2i(t->Sampler.MaxAnisotropy) << 27 |
nvgl_filter_mode(t->Sampler.MinFilter) << 24 |
(lod_bias & 0xff) << 16;
} else {
s = &to_nv04_context(ctx)->dummy_texture;
format |= NV04_TEXTURED_TRIANGLE_FORMAT_ADDRESSU_REPEAT |
NV04_TEXTURED_TRIANGLE_FORMAT_ADDRESSV_REPEAT |
1 << 12 |
NV04_TEXTURED_TRIANGLE_FORMAT_COLOR_A8R8G8B8;
filter |= NV04_TEXTURED_TRIANGLE_FILTER_MINIFY_NEAREST |
NV04_TEXTURED_TRIANGLE_FILTER_MAGNIFY_NEAREST;
}
nv04->texture[i] = s;
nv04->format[i] = format;
nv04->filter[i] = filter;
}
static Bool
setup_texture(NVPtr pNv, int unit, PicturePtr pict, PixmapPtr pixmap)
{
struct nouveau_channel *chan = pNv->chan;
struct nouveau_grobj *celsius = pNv->Nv3D;
struct nouveau_bo *bo = nouveau_pixmap_bo(pixmap);
unsigned tex_reloc = NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_RD;
long w = pict->pDrawable->width,
h = pict->pDrawable->height;
unsigned int txfmt =
NV10TCL_TX_FORMAT_WRAP_T_CLAMP_TO_EDGE |
NV10TCL_TX_FORMAT_WRAP_S_CLAMP_TO_EDGE |
log2i(w) << 20 | log2i(h) << 16 |
1 << 12 | /* lod == 1 */
get_tex_format(pict) |
0x50 /* UNK */;
BEGIN_RING(chan, celsius, NV10TCL_TX_OFFSET(unit), 1);
if (OUT_RELOCl(chan, bo, 0, tex_reloc))
return FALSE;
if (pict->repeat == RepeatNone) {
/* NPOT_SIZE expects an even number for width, we can
* round up uneven numbers here because EXA always
* gives 64 byte aligned pixmaps and for all formats
* we support 64 bytes represents an even number of
* pixels
*/
w = (w + 1) &~ 1;
BEGIN_RING(chan, celsius, NV10TCL_TX_NPOT_PITCH(unit), 1);
OUT_RING (chan, exaGetPixmapPitch(pixmap) << 16);
BEGIN_RING(chan, celsius, NV10TCL_TX_NPOT_SIZE(unit), 1);
OUT_RING (chan, w << 16 | h);
}
BEGIN_RING(chan, celsius, NV10TCL_TX_FORMAT(unit), 1 );
if (OUT_RELOCd(chan, bo, txfmt, tex_reloc | NOUVEAU_BO_OR,
NV10TCL_TX_FORMAT_DMA0, NV10TCL_TX_FORMAT_DMA1))
return FALSE;
BEGIN_RING(chan, celsius, NV10TCL_TX_ENABLE(unit), 1 );
OUT_RING (chan, NV10TCL_TX_ENABLE_ENABLE);
BEGIN_RING(chan, celsius, NV10TCL_TX_FILTER(unit), 1);
if (pict->filter == PictFilterNearest)
OUT_RING(chan, (NV10TCL_TX_FILTER_MAGNIFY_NEAREST |
NV10TCL_TX_FILTER_MINIFY_NEAREST));
else
OUT_RING(chan, (NV10TCL_TX_FILTER_MAGNIFY_LINEAR |
NV10TCL_TX_FILTER_MINIFY_LINEAR));
return TRUE;
}
static TIFF *tifopenr(const char *tif, // TIFF file name
bool *c, // return is complex?
int *l, // return log2 tile size
int *n, // return log2 height
int *m, // return log2 width
int *p) // return pixel size
{
TIFF *T;
if ((T = TIFFOpen(tif, "r")))
{
uint32 W = 0;
uint32 L = 0;
uint32 S = 0;
uint16 P = 0;
uint16 B = 0;
uint16 G = 0;
uint16 F = 0;
TIFFGetField(T, TIFFTAG_IMAGEWIDTH, &W);
TIFFGetField(T, TIFFTAG_IMAGELENGTH, &L);
TIFFGetField(T, TIFFTAG_TILEWIDTH, &S);
TIFFGetField(T, TIFFTAG_SAMPLESPERPIXEL, &P);
TIFFGetField(T, TIFFTAG_BITSPERSAMPLE, &B);
TIFFGetField(T, TIFFTAG_PLANARCONFIG, &G);
TIFFGetField(T, TIFFTAG_SAMPLEFORMAT, &F);
if (G == PLANARCONFIG_CONTIG)
{
if (F == SAMPLEFORMAT_IEEEFP)
{
if (B == 32)
{
if (ispow2(W) && ispow2(L))
{
*c = (P % 2) ? false : true;
*p = (P % 2) ? P : P / 2;
*l = log2i(S);
*n = log2i(L);
*m = log2i(W);
return T;
}
else apperr("TIFF image size must be power of 2");
}
else apperr("TIFF must have 32 bits per sample");
}
else apperr("TIFF must have floating point sample format");
}
else apperr("TIFF must have contiguous planar configuration");
TIFFClose(T);
}
return 0;
}
static Bool
NV30EXATexture(ScrnInfoPtr pScrn, PixmapPtr pPix, PicturePtr pPict, int unit)
{
NVPtr pNv = NVPTR(pScrn);
struct nouveau_channel *chan = pNv->chan;
struct nouveau_grobj *rankine = pNv->Nv3D;
struct nouveau_bo *bo = nouveau_pixmap_bo(pPix);
nv_pict_texture_format_t *fmt;
uint32_t card_filter, card_repeat;
uint32_t tex_reloc = NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_RD;
NV30EXA_STATE;
fmt = NV30_GetPictTextureFormat(pPict->format);
if (!fmt)
return FALSE;
card_repeat = 3; /* repeatNone */
if (pPict->filter == PictFilterBilinear)
card_filter = 2;
else
card_filter = 1;
BEGIN_RING(chan, rankine, NV34TCL_TX_OFFSET(unit), 8);
if (OUT_RELOCl(chan, bo, 0, tex_reloc) ||
OUT_RELOCd(chan, bo, NV34TCL_TX_FORMAT_DIMS_2D | (1 << 16) | 8 |
(fmt->card_fmt << NV34TCL_TX_FORMAT_FORMAT_SHIFT) |
(log2i(pPix->drawable.width) <<
NV34TCL_TX_FORMAT_BASE_SIZE_U_SHIFT) |
(log2i(pPix->drawable.height) <<
NV34TCL_TX_FORMAT_BASE_SIZE_V_SHIFT),
tex_reloc | NOUVEAU_BO_OR,
NV34TCL_TX_FORMAT_DMA0, NV34TCL_TX_FORMAT_DMA1))
return FALSE;
OUT_RING (chan, (card_repeat << NV34TCL_TX_WRAP_S_SHIFT) |
(card_repeat << NV34TCL_TX_WRAP_T_SHIFT) |
(card_repeat << NV34TCL_TX_WRAP_R_SHIFT));
OUT_RING (chan, NV34TCL_TX_ENABLE_ENABLE);
OUT_RING (chan, (((uint32_t)exaGetPixmapPitch(pPix)) << NV34TCL_TX_SWIZZLE_RECT_PITCH_SHIFT ) |
fmt->card_swz);
OUT_RING (chan, (card_filter << NV34TCL_TX_FILTER_MINIFY_SHIFT) /* min */ |
(card_filter << NV34TCL_TX_FILTER_MAGNIFY_SHIFT) /* mag */ |
0x2000 /* engine lock */);
OUT_RING (chan, (pPix->drawable.width << NV34TCL_TX_NPOT_SIZE_W_SHIFT) | pPix->drawable.height);
OUT_RING (chan, 0); /* border ARGB */
state->unit[unit].width = (float)pPix->drawable.width;
state->unit[unit].height = (float)pPix->drawable.height;
state->unit[unit].transform = pPict->transform;
return TRUE;
}
static int dm365_aew_open( AEW_WINDOW *window )
{
int bytes;
fd = open("/dev/dm365_aew", O_RDWR);
assert(fd > 0);
aew_config.window_config.hz_cnt = window->width / aew_config.window_config.width;
aew_config.window_config.vt_cnt = window->height / aew_config.window_config.height;
aew_config.window_config.hz_start = window->start_x;
aew_config.window_config.vt_start = window->start_y;
aew_config.blackwindow_config.height = window->start_y;
// 使颜色值范围为0-255
if(aew_config.alaw_enable == H3A_AEW_ENABLE)
{
aew_config.sum_shift = log2i(aew_config.window_config.width * aew_config.window_config.height / 4);
}
else
{
aew_config.sum_shift = log2i(aew_config.window_config.width * aew_config.window_config.height / 4) + 2;
}
bytes = ioctl(fd, AEW_S_PARAM, &aew_config);
if(bytes < 0)
{
perror("dm365_aew set param failed");
close(fd);
return -1;
}
else
{
printf("dm365_aew buffer size = %d\n", bytes);
}
aew_stat_buf = (unsigned short *)malloc(bytes);
if(aew_stat_buf == 0)
{
perror("dm365_aew malloc memory failed");
close(fd);
return -1;
}
aew_stat_buf_size = bytes;
aew_paxel_cnt = aew_config.window_config.hz_cnt * (aew_config.window_config.vt_cnt + 1);
return aew_paxel_cnt;
}
static void
nv20_fragtex_build(struct nv20_context *nv20, int unit)
{
#if 0
struct nv20_sampler_state *ps = nv20->tex_sampler[unit];
struct nv20_miptree *nv20mt = nv20->tex_miptree[unit];
struct pipe_texture *pt = &nv20mt->base;
struct nv20_texture_format *tf;
uint32_t txf, txs, txp;
tf = nv20_fragtex_format(pt->format);
if (!tf || !tf->defined) {
NOUVEAU_ERR("Unsupported texture format: 0x%x\n", pt->format);
return;
}
txf = tf->format << 8;
txf |= (pt->last_level + 1) << 16;
txf |= log2i(pt->width[0]) << 20;
txf |= log2i(pt->height[0]) << 24;
txf |= log2i(pt->depth[0]) << 28;
txf |= 8;
switch (pt->target) {
case PIPE_TEXTURE_CUBE:
txf |= NV10TCL_TX_FORMAT_CUBE_MAP;
/* fall-through */
case PIPE_TEXTURE_2D:
txf |= (2<<4);
break;
case PIPE_TEXTURE_1D:
txf |= (1<<4);
break;
default:
NOUVEAU_ERR("Unknown target %d\n", pt->target);
return;
}
BEGIN_RING(kelvin, NV10TCL_TX_OFFSET(unit), 8);
OUT_RELOCl(nv20mt->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_RD);
OUT_RELOCd(nv20mt->buffer,txf,NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_OR | NOUVEAU_BO_RD, 1/*VRAM*/,2/*TT*/);
OUT_RING (ps->wrap);
OUT_RING (0x40000000); /* enable */
OUT_RING (txs);
OUT_RING (ps->filt | 0x2000 /* magic */);
OUT_RING ((pt->width[0] << 16) | pt->height[0]);
OUT_RING (ps->bcol);
#endif
}
PageTable::PageTable()
: numOSPages("MemoryOS:numOSPages"),
numUsrPages("MemoryOS:numUsrPages")
{
invertedPT = new IntlPTEntry[MemoryOS::numPhysicalPages];
ulong tmp_numEntriesPage = GMemorySystem::getPageSize() / BytesPerEntry;
maskEntriesPage = tmp_numEntriesPage - 1;
log2EntriesPage = log2i(tmp_numEntriesPage);
/* We assume a 32 bit virtual address */
long numL1PTEntries = GMemorySystem::calcPage(0x80000000) >> (log2EntriesPage-1);
L1PT = new L1IntlPTEntry[numL1PTEntries];
// L1 Page Table
invertedPT[0].status = SystemPageStatus | ValidPageStatus;
for (ulong i = 1; i < MemoryOS::numPhysicalPages; i++)
invertedPT[i].status = 0;
for (long i = 0; i < numL1PTEntries; i++)
L1PT[i].status = 0;
}
MSHR<Addr_t, Cache_t>::MSHR(const char *name, int32_t size, int32_t lineSize, int32_t aPolicy)
:nEntries(size)
,Log2LineSize(log2i(lineSize))
,nUse("%s_MSHR:nUse", name)
,nUseReads("%s_MSHR:nUseReads", name)
,nUseWrites("%s_MSHR:nUseWrites", name)
,nOverflows("%s_MSHR:nOverflows", name)
,maxUsedEntries("%s_MSHR_maxUsedEntries", name)
,nCanAccept("%s_MSHR:nCanAccept", name)
,nCanNotAccept("%s_MSHR:nCanNotAccept", name)
,nCanNotAcceptConv("%s_MSHR:nCanNotAcceptConv", name)
,blockingCycles("%s_MSHR:blockingCycles",name)
,allocPolicy(aPolicy)
,occStatsAttached(false)
,lowerCache(NULL)
{
I(size>0 && size<1024*32);
#ifdef MSHR_BASICOCCSTATS
occStats = new MSHRStats<Addr_t,Cache_t>(name);
occStats->attach(this);
#endif
nFreeEntries = size;
}
static unsigned
get_swizzled_offset(struct nouveau_surface *s, unsigned x, unsigned y)
{
unsigned k = log2i(MIN2(s->width, s->height));
unsigned u = (x & 0x001) << 0 |
(x & 0x002) << 1 |
(x & 0x004) << 2 |
(x & 0x008) << 3 |
(x & 0x010) << 4 |
(x & 0x020) << 5 |
(x & 0x040) << 6 |
(x & 0x080) << 7 |
(x & 0x100) << 8 |
(x & 0x200) << 9 |
(x & 0x400) << 10 |
(x & 0x800) << 11;
unsigned v = (y & 0x001) << 1 |
(y & 0x002) << 2 |
(y & 0x004) << 3 |
(y & 0x008) << 4 |
(y & 0x010) << 5 |
(y & 0x020) << 6 |
(y & 0x040) << 7 |
(y & 0x080) << 8 |
(y & 0x100) << 9 |
(y & 0x200) << 10 |
(y & 0x400) << 11 |
(y & 0x800) << 12;
return s->cpp * (((u | v) & ~(~0 << 2*k)) |
(x & (~0 << k)) << k |
(y & (~0 << k)) << k);
}
/* static */
int UIWizardNewVDPage3::sizeMBToSlider(qulonglong uValue, int iSliderScale)
{
int iPower = log2i(uValue);
qulonglong uTickMB = qulonglong (1) << iPower;
qulonglong uTickMBNext = qulonglong (1) << (iPower + 1);
int iStep = (uValue - uTickMB) * iSliderScale / (uTickMBNext - uTickMB);
return iPower * iSliderScale + iStep;
}
static void test_log2i(void) {
assert_se(log2i(1) == 0);
assert_se(log2i(2) == 1);
assert_se(log2i(3) == 1);
assert_se(log2i(4) == 2);
assert_se(log2i(32) == 5);
assert_se(log2i(33) == 5);
assert_se(log2i(63) == 5);
assert_se(log2i(INT_MAX) == sizeof(int)*8-2);
}
static int reorder_frame_offset(int idx, int sub_gop)
{
#if DYADIC_CODING
return dyadic_reorder_code_to_display[log2i(sub_gop)][idx]-sub_gop+1;
#else
if (idx==0) return 0;
else return idx-sub_gop;
#endif
}
static int reorder_frame_offset(int idx, int sub_gop, int dyadic)
{
if (dyadic && sub_gop>1) {
return dyadic_reorder_code_to_display[log2i(sub_gop)][idx]-sub_gop+1;
} else {
if (idx==0) return 0;
else return idx-sub_gop;
}
}
void inverse_transform (const int16_t * coeff, int16_t *block, int size)
{
if (size < 32)
inverse_transform_simd(coeff, block, size);
else {
int i,j;
int out[MAX_TR_SIZE];
int16_t tmp[MAX_TR_SIZE*MAX_TR_SIZE];
int tr = log2i(size)-2;
const int16_t * tr_matrix = transform_table[tr];
const int shift_1 = 7;
const int add_1 = 1 << (shift_1 - 1);
const int shift_2 = 12;
const int add_2 = 1 << (shift_2 -1);
/* 1st dimension */
for (i=0;i<MAX_QUANT_SIZE;i++){
#if 1
/* Partial factorization */
transform_1d_even_l0(coeff, tr_matrix, size, i, out);
#else
/* Matrix multiplication */
int k;
for (j=0;j<size;j++){
out[j] = 0;
for (k=0;k<MAX_QUANT_SIZE;k++){
out[j] += tr_matrix[k*size+j] * coeff[k*size+i];
}
}
#endif
for (j=0;j<size;j++){
tmp[i*size+j] = clip((out[j] + add_1) >> shift_1, -32768, 32767);
}
}
/* 2nd dimension */
for (i=0;i<size;i++){
#if 1
/* Partial factorization */
transform_1d_even_l0(tmp, tr_matrix, size, i, out);
#else
/* Matrix multiplication */
int k;
for (j=0;j<size;j++){
out[j] = 0;
for (k=0;k<MAX_QUANT_SIZE;k++){
out[j] += tr_matrix[k*size+j] * tmp[k*size+i];
}
}
#endif
for (j=0;j<size;j++){
block[i*size + j] = clip((out[j] + add_2) >> shift_2, -32768, 32767);
}
}
}
}
static void test_log2i(void) {
log_info("/* %s */", __func__);
assert_se(log2i(1) == 0);
assert_se(log2i(2) == 1);
assert_se(log2i(3) == 1);
assert_se(log2i(4) == 2);
assert_se(log2i(32) == 5);
assert_se(log2i(33) == 5);
assert_se(log2i(63) == 5);
assert_se(log2i(INT_MAX) == sizeof(int)*8-2);
}
LiveCache::LiveCache() {
cacheBank = CacheType::create("LiveCache", "", "LL");
lineSize = cacheBank->getLineSize();
lineSizeBits = log2i(lineSize);
I(getLineSize() < 4096); // To avoid bank selection conflict (insane LiveCache line)
lineCount = (uint64_t) cacheBank->getNumLines();
maxOrder = 0;
}
void decode_and_reconstruct_block_intra (uint8_t *rec, int stride, int size, int qp, uint8_t *pblock, int16_t *coeffq,
int tb_split, int upright_available,int downleft_available, intra_mode_t intra_mode,int ypos,int xpos,int width,int comp,
qmtx_t ** iwmatrix){
int16_t *rcoeff = thor_alloc(2*MAX_TR_SIZE*MAX_TR_SIZE, 16);
int16_t *rblock = thor_alloc(2*MAX_TR_SIZE*MAX_TR_SIZE, 16);
int16_t *rblock2 = thor_alloc(2*MAX_TR_SIZE*MAX_TR_SIZE, 16);
uint8_t* left_data = (uint8_t*)thor_alloc(2*MAX_TR_SIZE+2,16)+1;
uint8_t* top_data = (uint8_t*)thor_alloc(2*MAX_TR_SIZE+2,16)+1;
uint8_t top_left;
if (tb_split){
int size2 = size/2;
int i,j,index;
for (i=0;i<size;i+=size2){
for (j=0;j<size;j+=size2){
make_top_and_left(left_data,top_data,&top_left,rec,stride,&rec[i*stride+j],stride,i,j,ypos,xpos,size2,upright_available,downleft_available,1);
get_intra_prediction(left_data,top_data,top_left,ypos+i,xpos+j,size2,pblock,intra_mode);
index = 2*(i/size2) + (j/size2);
dequantize (coeffq+index*size2*size2, rcoeff, qp, size2, iwmatrix ? iwmatrix[log2i(size2/4)] : NULL, MAX_QUANT_SIZE);
inverse_transform (rcoeff, rblock2, size2);
reconstruct_block(rblock2,pblock,&rec[i*stride+j],size2,stride);
}
}
}
else{
make_top_and_left(left_data,top_data,&top_left,rec,stride,NULL,0,0,0,ypos,xpos,size,upright_available,downleft_available,0);
get_intra_prediction(left_data,top_data,top_left,ypos,xpos,size,pblock,intra_mode);
dequantize (coeffq, rcoeff, qp, size, iwmatrix ? iwmatrix[log2i(size/4)] : NULL, MAX_QUANT_SIZE);
inverse_transform (rcoeff, rblock, size);
reconstruct_block(rblock,pblock,rec,size,stride);
}
thor_free(top_data - 1);
thor_free(left_data - 1);
thor_free(rcoeff);
thor_free(rblock);
thor_free(rblock2);
}
void STree_Grow( STreeNode *pseudo, int nNodes ) {
int nLeaves, nSpineNodes;
nLeaves = nNodes + 1 - ( 1 << log2i( nNodes + 1 ) );
STree_Compression( pseudo, nLeaves );
nSpineNodes = nNodes - nLeaves;
while ( nSpineNodes > 1 ) {
nSpineNodes /= 2;
STree_Compression( pseudo, nSpineNodes );
}
}
void check_parameters(enc_params *params)
{
if(params->num_frames <= 0)
{
fatalerror("Number of frames must be positive");
}
if (params->width%8 || params->height%8)
{
fatalerror("Width and height must be a multiple of 8\n");
}
if (params->max_num_ref < 1 || params->max_num_ref > 4)
{
fatalerror("This number of max reference frames is not supported \n");
}
if(params->max_delta_qp >= 8)
{
fatalerror("max_delta_qp too large\n");
}
if(params->HQperiod >= MAX_REF_FRAMES)
{
fatalerror("HQperiod too large");
}
if (params->num_reorder_pics > 0 && params->HQperiod > 1 && (params->HQperiod % (params->num_reorder_pics+1))!=0) {
fatalerror("Subgop length (num_reorder_pics+1) must divide HQperiod.\n");
}
if (params->dyadic_coding) {
int nrp1 = params->num_reorder_pics+1;
if (nrp1 != (1<<(log2i(nrp1)))) {
fatalerror("num_reorder_pics+1 must be a power of 2 with dyadic coding.\n");
}
}
if (params->num_reorder_pics > 0 && params->max_num_ref < 2) {
fatalerror("More than one reference frame required for reordered pictures.\n");
}
if (params->intra_period % (params->num_reorder_pics+1)) {
fatalerror("Intra period must be a multiple of the subgroup size (num_reorder_pics+1).\n");
}
if (params->sync && params->encoder_speed<2) {
fatalerror("Sync requires encoder_speed=2\n");
}
if (params->bitrate > 0 && params->num_reorder_pics > 0){
fatalerror("Current rate control doesn't work with frame reordering\n");
}
}
/*
* Estimate the number of bisect steps left (after the current step)
*
* For any x between 0 included and 2^n excluded, the probability for
* n - 1 steps left looks like:
*
* P(2^n + x) == (2^n - x) / (2^n + x)
*
* and P(2^n + x) < 0.5 means 2^n < 3x
*/
int estimate_bisect_steps(int all)
{
int n, x, e;
if (all < 3)
return 0;
n = log2i(all);
e = exp2i(n);
x = all - e;
return (e < 3 * x) ? n : n - 1;
}
IUnknown* DX11Engine::createTexture2D_SRV(IUnknown* pDeviceIn, const Texture& info,
const char* buffer, size_t size, const char* name) const
{
auto pDevice = reinterpret_cast<ID3D11Device*>(pDeviceIn);
HRESULT hr;
D3D11_TEXTURE2D_DESC tex = {
uint32_t(info.mWidth), uint32_t(info.mHeight),
uint32_t(info.mMipCount), uint32_t(info.mArraySize), info.mFormat, { 1, 0 },
D3D11_USAGE_DEFAULT, D3D11_BIND_SHADER_RESOURCE,
0, 0
};
if(info.eMiscGenerateMipMaps) {
tex.MipLevels = 0;
tex.MiscFlags |= D3D11_RESOURCE_MISC_GENERATE_MIPS;
tex.BindFlags |= D3D11_BIND_RENDER_TARGET;
}
uint32_t bpp = getBitsPerPixel(info.mFormat);
int level = 1;
if (info.eMiscGenerateMipMaps)
level = log2i(std::max(info.mWidth, info.mHeight)) + 1;
boost::container::small_vector<D3D11_SUBRESOURCE_DATA, 12> subres(level);
int w = info.mWidth;
int h = info.mHeight;
for (int i = 0; i != level; ++i) {
subres[i] = D3D11_SUBRESOURCE_DATA{ buffer, w * bpp / 8, 0 };
buffer += mipmap_size<1>(w, h, bpp);
w = next_mip_size(w);
h = next_mip_size(h);
}
std::unique_ptr<ID3D11Texture2D, COMDeleter> pTex;
std::unique_ptr<ID3D11ShaderResourceView, COMDeleter> pSRV;
CD3D11_SHADER_RESOURCE_VIEW_DESC srvDesc(D3D11_SRV_DIMENSION_TEXTURE2D);
V(pDevice->CreateTexture2D(&tex, subres.data(), ref(pTex)));
V(pDevice->CreateShaderResourceView(pTex.get(), &srvDesc, ref(pSRV)));
STAR_SET_DEBUG_NAME(pSRV, size, name);
return pSRV.release();
}
/* static */
int UIWizardNewVDPage3::calculateSliderScale(qulonglong uMaximumMediumSize)
{
/* Detect how many steps to recognize between adjacent powers of 2
* to ensure that the last slider step is exactly that we need: */
int iSliderScale = 0;
int iPower = log2i(uMaximumMediumSize);
qulonglong uTickMB = (qulonglong)1 << iPower;
if (uTickMB < uMaximumMediumSize)
{
qulonglong uTickMBNext = (qulonglong)1 << (iPower + 1);
qulonglong uGap = uTickMBNext - uMaximumMediumSize;
iSliderScale = (int)((uTickMBNext - uTickMB) / uGap);
}
return qMax(iSliderScale, 8);
}
void configure() {
if (m_nested == NULL)
Log(EError, "A nested volume data source is needed!");
m_aabb = m_nested->getAABB();
if (!m_aabb.isValid())
Log(EError, "Nested axis-aligned bounding box was invalid!");
if (m_voxelWidth == -1)
m_voxelWidth = m_nested->getStepSize();
size_t memoryLimitPerCore = m_memoryLimit
/ std::max((size_t) 1, Scheduler::getInstance()->getLocalWorkerCount());
Vector totalCells = m_aabb.getExtents() / m_voxelWidth;
for (int i=0; i<3; ++i)
m_cellCount[i] = (int) std::ceil(totalCells[i]);
if (m_nested->supportsFloatLookups())
m_channels = 1;
else if (m_nested->supportsVectorLookups())
m_channels = 1;
else if (m_nested->supportsSpectrumLookups())
m_channels = SPECTRUM_SAMPLES;
else
Log(EError, "Nested volume offers no access methods!");
m_blockRes = m_blockSize+1;
int blockMemoryUsage = (int) std::pow((Float) m_blockRes, 3) * m_channels * sizeof(float);
m_blocksPerCore = memoryLimitPerCore / blockMemoryUsage;
m_worldToVolume = m_volumeToWorld.inverse();
m_worldToGrid = Transform::scale(Vector(1/m_voxelWidth))
* Transform::translate(-Vector(m_aabb.min)) * m_worldToVolume;
m_voxelMask = m_blockSize-1;
m_blockMask = ~(m_blockSize-1);
m_blockShift = log2i((uint32_t) m_blockSize);
Log(EInfo, "Volume cache configuration");
Log(EInfo, " Block size in voxels = %i", m_blockSize);
Log(EInfo, " Voxel width = %f", m_voxelWidth);
Log(EInfo, " Memory usage of one block = %s", memString(blockMemoryUsage).c_str());
Log(EInfo, " Memory limit = %s", memString(m_memoryLimit).c_str());
Log(EInfo, " Memory limit per core = %s", memString(memoryLimitPerCore).c_str());
Log(EInfo, " Max. blocks per core = %i", m_blocksPerCore);
Log(EInfo, " Effective resolution = %s", totalCells.toString().c_str());
Log(EInfo, " Effective storage = %s", memString((size_t)
(totalCells[0]*totalCells[1]*totalCells[2]*sizeof(float)*m_channels)).c_str());
}
/*************************************************************************
* This function is the entry point for KWMETIS
**************************************************************************/
void METIS_WPartGraphKway(int *nvtxs, idxtype *xadj, idxtype *adjncy, idxtype *vwgt,
idxtype *adjwgt, int *wgtflag, int *numflag, int *nparts,
float *tpwgts, int *options, int *edgecut, idxtype *part)
{
int i, j;
GraphType graph;
CtrlType ctrl;
if (*numflag == 1)
Change2CNumbering(*nvtxs, xadj, adjncy);
SetUpGraph(&graph, OP_KMETIS, *nvtxs, 1, xadj, adjncy, vwgt, adjwgt, *wgtflag);
if (options[0] == 0) { /* Use the default parameters */
ctrl.CType = KMETIS_CTYPE;
ctrl.IType = KMETIS_ITYPE;
ctrl.RType = KMETIS_RTYPE;
ctrl.dbglvl = KMETIS_DBGLVL;
}
else {
ctrl.CType = options[OPTION_CTYPE];
ctrl.IType = options[OPTION_ITYPE];
ctrl.RType = options[OPTION_RTYPE];
ctrl.dbglvl = options[OPTION_DBGLVL];
}
ctrl.optype = OP_KMETIS;
ctrl.CoarsenTo = amax((*nvtxs)/(40*log2i(*nparts)), 20*(*nparts));
ctrl.maxvwgt = 1.5*((graph.vwgt ? idxsum(*nvtxs, graph.vwgt) : (*nvtxs))/ctrl.CoarsenTo);
InitRandom(-1);
AllocateWorkSpace(&ctrl, &graph, *nparts);
IFSET(ctrl.dbglvl, DBG_TIME, InitTimers(&ctrl));
IFSET(ctrl.dbglvl, DBG_TIME, starttimer(ctrl.TotalTmr));
*edgecut = MlevelKWayPartitioning(&ctrl, &graph, *nparts, part, tpwgts, 1.03);
IFSET(ctrl.dbglvl, DBG_TIME, stoptimer(ctrl.TotalTmr));
IFSET(ctrl.dbglvl, DBG_TIME, PrintTimers(&ctrl));
FreeWorkSpace(&ctrl, &graph);
if (*numflag == 1)
Change2FNumbering(*nvtxs, xadj, adjncy, part);
}
void preproc_emit_line_info(int lineno, const char *fname, enum lineinfo lineinfo)
{
unsigned lineinfobits = lineinfo;
/* output PP info */
if(no_output || !option_line_info)
return;
printf("# %d \"%s\"", lineno, fname);
while(lineinfobits){
unsigned bit = extractbottombit(&lineinfobits);
int n = log2i(bit);
printf(" %d", n);
}
putchar('\n');
}
sexp sexp_write_bignum (sexp ctx, sexp a, sexp out, sexp_uint_t base) {
int i, str_len, lg_base = log2i(base);
char *data;
sexp_gc_var2(b, str);
sexp_gc_preserve2(ctx, b, str);
b = sexp_copy_bignum(ctx, NULL, a, 0);
sexp_bignum_sign(b) = 1;
i = str_len = (sexp_bignum_length(b)*sizeof(sexp_uint_t)*8 + lg_base - 1)
/ lg_base + 1;
str = sexp_make_string(ctx, sexp_make_fixnum(str_len),
sexp_make_character(' '));
data = sexp_string_data(str);
while (! sexp_bignum_zerop(b))
data[--i] = hex_digit(sexp_bignum_fxdiv(ctx, b, base, 0));
if (i == str_len)
data[--i] = '0';
else if (sexp_bignum_sign(a) == -1)
data[--i] = '-';
sexp_write_string(ctx, data + i, out);
sexp_gc_release2(ctx);
return SEXP_VOID;
}
void setFilmResolution(const Vector2i &res, bool blocked) {
if (blocked) {
/* Determine parameters of the space partition in the first two
dimensions. This is required to support blocked rendering. */
for (int i=0; i<2; ++i)
m_resolution[i] = std::min((uint32_t) MAX_RESOLUTION,
roundToPowerOfTwo((uint32_t) res[i]));
m_logHeight = log2i((uint32_t) m_resolution.y);
m_samplesPerBatch = m_sampleCount;
m_factor = (Float) 1.0f / (m_sampleCount *
(size_t) m_resolution.x * (size_t) m_resolution.y);
m_offset = 0;
m_stride = m_resolution.y;
} else {
m_samplesPerBatch = m_sampleCount *
(size_t) res.x * (size_t) res.y;
m_factor = (Float) 1.0f / m_samplesPerBatch;
m_resolution = Vector2i(1);
m_offset = 0;
m_stride = 1;
}
m_pixelPosition = Point2i(0);
}
static ALCenum oss_open_capture(ALCdevice *device, const ALCchar *deviceName)
{
int numFragmentsLogSize;
int log2FragmentSize;
unsigned int periods;
audio_buf_info info;
ALuint frameSize;
int numChannels;
oss_data *data;
int ossFormat;
int ossSpeed;
char *err;
if(!deviceName)
deviceName = oss_device;
else if(strcmp(deviceName, oss_device) != 0)
return ALC_INVALID_VALUE;
data = (oss_data*)calloc(1, sizeof(oss_data));
data->killNow = 0;
data->fd = open(oss_capture, O_RDONLY);
if(data->fd == -1)
{
free(data);
ERR("Could not open %s: %s\n", oss_capture, strerror(errno));
return ALC_INVALID_VALUE;
}
switch(device->FmtType)
{
case DevFmtByte:
ossFormat = AFMT_S8;
break;
case DevFmtUByte:
ossFormat = AFMT_U8;
break;
case DevFmtShort:
ossFormat = AFMT_S16_NE;
break;
case DevFmtUShort:
case DevFmtInt:
case DevFmtUInt:
case DevFmtFloat:
free(data);
ERR("%s capture samples not supported\n", DevFmtTypeString(device->FmtType));
return ALC_INVALID_VALUE;
}
periods = 4;
numChannels = ChannelsFromDevFmt(device->FmtChans);
frameSize = numChannels * BytesFromDevFmt(device->FmtType);
ossSpeed = device->Frequency;
log2FragmentSize = log2i(device->UpdateSize * device->NumUpdates *
frameSize / periods);
/* according to the OSS spec, 16 bytes are the minimum */
if (log2FragmentSize < 4)
log2FragmentSize = 4;
numFragmentsLogSize = (periods << 16) | log2FragmentSize;
#define CHECKERR(func) if((func) < 0) { \
err = #func; \
goto err; \
}
CHECKERR(ioctl(data->fd, SNDCTL_DSP_SETFRAGMENT, &numFragmentsLogSize));
CHECKERR(ioctl(data->fd, SNDCTL_DSP_SETFMT, &ossFormat));
CHECKERR(ioctl(data->fd, SNDCTL_DSP_CHANNELS, &numChannels));
CHECKERR(ioctl(data->fd, SNDCTL_DSP_SPEED, &ossSpeed));
CHECKERR(ioctl(data->fd, SNDCTL_DSP_GETISPACE, &info));
if(0)
{
err:
ERR("%s failed: %s\n", err, strerror(errno));
close(data->fd);
free(data);
return ALC_INVALID_VALUE;
}
#undef CHECKERR
if((int)ChannelsFromDevFmt(device->FmtChans) != numChannels)
{
ERR("Failed to set %s, got %d channels instead\n", DevFmtChannelsString(device->FmtChans), numChannels);
close(data->fd);
free(data);
return ALC_INVALID_VALUE;
}
if(!((ossFormat == AFMT_S8 && device->FmtType == DevFmtByte) ||
(ossFormat == AFMT_U8 && device->FmtType == DevFmtUByte) ||
(ossFormat == AFMT_S16_NE && device->FmtType == DevFmtShort)))
{
ERR("Failed to set %s samples, got OSS format %#x\n", DevFmtTypeString(device->FmtType), ossFormat);
close(data->fd);
free(data);
return ALC_INVALID_VALUE;
}
data->ring = CreateRingBuffer(frameSize, device->UpdateSize * device->NumUpdates);
if(!data->ring)
{
ERR("Ring buffer create failed\n");
close(data->fd);
free(data);
return ALC_OUT_OF_MEMORY;
}
data->data_size = info.fragsize;
data->mix_data = calloc(1, data->data_size);
device->ExtraData = data;
data->thread = StartThread(OSSCaptureProc, device);
if(data->thread == NULL)
{
device->ExtraData = NULL;
free(data->mix_data);
free(data);
return ALC_OUT_OF_MEMORY;
}
device->szDeviceName = strdup(deviceName);
return ALC_NO_ERROR;
}
static ALCboolean oss_reset_playback(ALCdevice *device)
{
oss_data *data = (oss_data*)device->ExtraData;
int numFragmentsLogSize;
int log2FragmentSize;
unsigned int periods;
audio_buf_info info;
ALuint frameSize;
int numChannels;
int ossFormat;
int ossSpeed;
char *err;
switch(device->FmtType)
{
case DevFmtByte:
ossFormat = AFMT_S8;
break;
case DevFmtUByte:
ossFormat = AFMT_U8;
break;
case DevFmtUShort:
case DevFmtInt:
case DevFmtUInt:
case DevFmtFloat:
device->FmtType = DevFmtShort;
/* fall-through */
case DevFmtShort:
ossFormat = AFMT_S16_NE;
break;
}
periods = device->NumUpdates;
numChannels = ChannelsFromDevFmt(device->FmtChans);
frameSize = numChannels * BytesFromDevFmt(device->FmtType);
ossSpeed = device->Frequency;
log2FragmentSize = log2i(device->UpdateSize * frameSize);
/* according to the OSS spec, 16 bytes are the minimum */
if (log2FragmentSize < 4)
log2FragmentSize = 4;
/* Subtract one period since the temp mixing buffer counts as one. Still
* need at least two on the card, though. */
if(periods > 2) periods--;
numFragmentsLogSize = (periods << 16) | log2FragmentSize;
#define CHECKERR(func) if((func) < 0) { \
err = #func; \
goto err; \
}
/* Don't fail if SETFRAGMENT fails. We can handle just about anything
* that's reported back via GETOSPACE */
ioctl(data->fd, SNDCTL_DSP_SETFRAGMENT, &numFragmentsLogSize);
CHECKERR(ioctl(data->fd, SNDCTL_DSP_SETFMT, &ossFormat));
CHECKERR(ioctl(data->fd, SNDCTL_DSP_CHANNELS, &numChannels));
CHECKERR(ioctl(data->fd, SNDCTL_DSP_SPEED, &ossSpeed));
CHECKERR(ioctl(data->fd, SNDCTL_DSP_GETOSPACE, &info));
if(0)
{
err:
ERR("%s failed: %s\n", err, strerror(errno));
return ALC_FALSE;
}
#undef CHECKERR
if((int)ChannelsFromDevFmt(device->FmtChans) != numChannels)
{
ERR("Failed to set %s, got %d channels instead\n", DevFmtChannelsString(device->FmtChans), numChannels);
return ALC_FALSE;
}
if(!((ossFormat == AFMT_S8 && device->FmtType == DevFmtByte) ||
(ossFormat == AFMT_U8 && device->FmtType == DevFmtUByte) ||
(ossFormat == AFMT_S16_NE && device->FmtType == DevFmtShort)))
{
ERR("Failed to set %s samples, got OSS format %#x\n", DevFmtTypeString(device->FmtType), ossFormat);
return ALC_FALSE;
}
device->Frequency = ossSpeed;
device->UpdateSize = info.fragsize / frameSize;
device->NumUpdates = info.fragments + 1;
SetDefaultChannelOrder(device);
return ALC_TRUE;
}
