static void
nve4_m2mf_copy_linear(struct nouveau_context *nv,
struct nouveau_bo *dst, unsigned dstoff, unsigned dstdom,
struct nouveau_bo *src, unsigned srcoff, unsigned srcdom,
unsigned size)
{
struct nouveau_pushbuf *push = nv->pushbuf;
struct nouveau_bufctx *bctx = nvc0_context(&nv->pipe)->bufctx;
nouveau_bufctx_refn(bctx, 0, src, srcdom | NOUVEAU_BO_RD);
nouveau_bufctx_refn(bctx, 0, dst, dstdom | NOUVEAU_BO_WR);
nouveau_pushbuf_bufctx(push, bctx);
nouveau_pushbuf_validate(push);
BEGIN_NVC0(push, SUBC_COPY(0x0400), 4);
PUSH_DATAh(push, src->offset + srcoff);
PUSH_DATA (push, src->offset + srcoff);
PUSH_DATAh(push, dst->offset + dstoff);
PUSH_DATA (push, dst->offset + dstoff);
BEGIN_NVC0(push, SUBC_COPY(0x0418), 1);
PUSH_DATA (push, size);
BEGIN_NVC0(push, SUBC_COPY(0x0300), 1);
PUSH_DATA (push, 0x186);
nouveau_bufctx_reset(bctx, 0);
}
static void
nv50_compute_upload_input(struct nv50_context *nv50, const uint32_t *input)
{
struct nv50_screen *screen = nv50->screen;
struct nouveau_pushbuf *push = screen->base.pushbuf;
unsigned size = align(nv50->compprog->parm_size, 0x4);
BEGIN_NV04(push, NV50_COMPUTE(USER_PARAM_COUNT), 1);
PUSH_DATA (push, (size / 4) << 8);
if (size) {
struct nouveau_mm_allocation *mm;
struct nouveau_bo *bo = NULL;
unsigned offset;
mm = nouveau_mm_allocate(screen->base.mm_GART, size, &bo, &offset);
assert(mm);
nouveau_bo_map(bo, 0, screen->base.client);
memcpy(bo->map + offset, input, size);
nouveau_bufctx_refn(nv50->bufctx, 0, bo, NOUVEAU_BO_GART | NOUVEAU_BO_RD);
nouveau_pushbuf_bufctx(push, nv50->bufctx);
nouveau_pushbuf_validate(push);
BEGIN_NV04(push, NV50_COMPUTE(USER_PARAM(0)), size / 4);
nouveau_pushbuf_data(push, bo, offset, size);
nouveau_fence_work(screen->base.fence.current, nouveau_mm_free_work, mm);
nouveau_bo_ref(NULL, &bo);
nouveau_bufctx_reset(nv50->bufctx, 0);
}
}
static void
nvc0_m2mf_copy_linear(struct nouveau_context *nv,
struct nouveau_bo *dst, unsigned dstoff, unsigned dstdom,
struct nouveau_bo *src, unsigned srcoff, unsigned srcdom,
unsigned size)
{
struct nouveau_pushbuf *push = nv->pushbuf;
struct nouveau_bufctx *bctx = nvc0_context(&nv->pipe)->bufctx;
nouveau_bufctx_refn(bctx, 0, src, srcdom | NOUVEAU_BO_RD);
nouveau_bufctx_refn(bctx, 0, dst, dstdom | NOUVEAU_BO_WR);
nouveau_pushbuf_bufctx(push, bctx);
nouveau_pushbuf_validate(push);
while (size) {
unsigned bytes = MIN2(size, 1 << 17);
BEGIN_NVC0(push, NVC0_M2MF(OFFSET_OUT_HIGH), 2);
PUSH_DATAh(push, dst->offset + dstoff);
PUSH_DATA (push, dst->offset + dstoff);
BEGIN_NVC0(push, NVC0_M2MF(OFFSET_IN_HIGH), 2);
PUSH_DATAh(push, src->offset + srcoff);
PUSH_DATA (push, src->offset + srcoff);
BEGIN_NVC0(push, NVC0_M2MF(LINE_LENGTH_IN), 2);
PUSH_DATA (push, bytes);
PUSH_DATA (push, 1);
BEGIN_NVC0(push, NVC0_M2MF(EXEC), 1);
PUSH_DATA (push, (1 << NVC0_M2MF_EXEC_INC__SHIFT) |
NVC0_M2MF_EXEC_LINEAR_IN | NVC0_M2MF_EXEC_LINEAR_OUT);
srcoff += bytes;
dstoff += bytes;
size -= bytes;
}
nouveau_bufctx_reset(bctx, 0);
}
void
nvc0_m2mf_push_linear(struct nouveau_context *nv,
struct nouveau_bo *dst, unsigned offset, unsigned domain,
unsigned size, const void *data)
{
struct nvc0_context *nvc0 = nvc0_context(&nv->pipe);
struct nouveau_pushbuf *push = nv->pushbuf;
uint32_t *src = (uint32_t *)data;
unsigned count = (size + 3) / 4;
nouveau_bufctx_refn(nvc0->bufctx, 0, dst, domain | NOUVEAU_BO_WR);
nouveau_pushbuf_bufctx(push, nvc0->bufctx);
nouveau_pushbuf_validate(push);
while (count) {
unsigned nr;
if (!PUSH_SPACE(push, 16))
break;
nr = PUSH_AVAIL(push);
assert(nr >= 16);
nr = MIN2(count, nr - 9);
nr = MIN2(nr, NV04_PFIFO_MAX_PACKET_LEN);
BEGIN_NVC0(push, NVC0_M2MF(OFFSET_OUT_HIGH), 2);
PUSH_DATAh(push, dst->offset + offset);
PUSH_DATA (push, dst->offset + offset);
BEGIN_NVC0(push, NVC0_M2MF(LINE_LENGTH_IN), 2);
PUSH_DATA (push, MIN2(size, nr * 4));
PUSH_DATA (push, 1);
BEGIN_NVC0(push, NVC0_M2MF(EXEC), 1);
PUSH_DATA (push, 0x100111);
/* must not be interrupted (trap on QUERY fence, 0x50 works however) */
BEGIN_NIC0(push, NVC0_M2MF(DATA), nr);
PUSH_DATAp(push, src, nr);
count -= nr;
src += nr;
offset += nr * 4;
size -= nr * 4;
}
nouveau_bufctx_reset(nvc0->bufctx, 0);
}
struct nouveau_bufref *
nouveau_bufctx_mthd(struct nouveau_bufctx *bctx, int bin, uint32_t packet,
struct nouveau_bo *bo, uint64_t data, uint32_t flags,
uint32_t vor, uint32_t tor)
{
struct nouveau_bufctx_priv *pctx = nouveau_bufctx(bctx);
struct nouveau_bufbin_priv *pbin = &pctx->bins[bin];
struct nouveau_bufref *bref = nouveau_bufctx_refn(bctx, bin, bo, flags);
if (bref) {
bref->packet = packet;
bref->data = data;
bref->vor = vor;
bref->tor = tor;
pbin->relocs++;
bctx->relocs++;
}
return bref;
}
void
nvc0_cb_push(struct nouveau_context *nv,
struct nouveau_bo *bo, unsigned domain,
unsigned base, unsigned size,
unsigned offset, unsigned words, const uint32_t *data)
{
struct nouveau_bufctx *bctx = nvc0_context(&nv->pipe)->bufctx;
struct nouveau_pushbuf *push = nv->pushbuf;
NOUVEAU_DRV_STAT(nv->screen, constbuf_upload_count, 1);
NOUVEAU_DRV_STAT(nv->screen, constbuf_upload_bytes, words * 4);
assert(!(offset & 3));
size = align(size, 0x100);
nouveau_bufctx_refn(bctx, 0, bo, NOUVEAU_BO_WR | domain);
nouveau_pushbuf_bufctx(push, bctx);
nouveau_pushbuf_validate(push);
BEGIN_NVC0(push, NVC0_3D(CB_SIZE), 3);
PUSH_DATA (push, size);
PUSH_DATAh(push, bo->offset + base);
PUSH_DATA (push, bo->offset + base);
while (words) {
unsigned nr = PUSH_AVAIL(push);
nr = MIN2(nr, words);
nr = MIN2(nr, NV04_PFIFO_MAX_PACKET_LEN - 1);
BEGIN_1IC0(push, NVC0_3D(CB_POS), nr + 1);
PUSH_DATA (push, offset);
PUSH_DATAp(push, data, nr);
words -= nr;
data += nr;
offset += nr * 4;
}
nouveau_bufctx_reset(bctx, 0);
}
void
nve4_p2mf_push_linear(struct nouveau_context *nv,
struct nouveau_bo *dst, unsigned offset, unsigned domain,
unsigned size, const void *data)
{
struct nvc0_context *nvc0 = nvc0_context(&nv->pipe);
struct nouveau_pushbuf *push = nv->pushbuf;
uint32_t *src = (uint32_t *)data;
unsigned count = (size + 3) / 4;
nouveau_bufctx_refn(nvc0->bufctx, 0, dst, domain | NOUVEAU_BO_WR);
nouveau_pushbuf_bufctx(push, nvc0->bufctx);
nouveau_pushbuf_validate(push);
while (count) {
unsigned nr = MIN2(count, (NV04_PFIFO_MAX_PACKET_LEN - 1));
if (!PUSH_SPACE(push, nr + 10))
break;
BEGIN_NVC0(push, NVE4_P2MF(UPLOAD_DST_ADDRESS_HIGH), 2);
PUSH_DATAh(push, dst->offset + offset);
PUSH_DATA (push, dst->offset + offset);
BEGIN_NVC0(push, NVE4_P2MF(UPLOAD_LINE_LENGTH_IN), 2);
PUSH_DATA (push, MIN2(size, nr * 4));
PUSH_DATA (push, 1);
/* must not be interrupted (trap on QUERY fence, 0x50 works however) */
BEGIN_1IC0(push, NVE4_P2MF(UPLOAD_EXEC), nr + 1);
PUSH_DATA (push, 0x1001);
PUSH_DATAp(push, src, nr);
count -= nr;
src += nr;
offset += nr * 4;
size -= nr * 4;
}
nouveau_bufctx_reset(nvc0->bufctx, 0);
}
static void
nvc0_m2mf_transfer_rect(struct nvc0_context *nvc0,
const struct nv50_m2mf_rect *dst,
const struct nv50_m2mf_rect *src,
uint32_t nblocksx, uint32_t nblocksy)
{
struct nouveau_pushbuf *push = nvc0->base.pushbuf;
struct nouveau_bufctx *bctx = nvc0->bufctx;
const int cpp = dst->cpp;
uint32_t src_ofst = src->base;
uint32_t dst_ofst = dst->base;
uint32_t height = nblocksy;
uint32_t sy = src->y;
uint32_t dy = dst->y;
uint32_t exec = (1 << 20);
assert(dst->cpp == src->cpp);
nouveau_bufctx_refn(bctx, 0, src->bo, src->domain | NOUVEAU_BO_RD);
nouveau_bufctx_refn(bctx, 0, dst->bo, dst->domain | NOUVEAU_BO_WR);
nouveau_pushbuf_bufctx(push, bctx);
nouveau_pushbuf_validate(push);
if (nouveau_bo_memtype(src->bo)) {
BEGIN_NVC0(push, NVC0_M2MF(TILING_MODE_IN), 5);
PUSH_DATA (push, src->tile_mode);
PUSH_DATA (push, src->width * cpp);
PUSH_DATA (push, src->height);
PUSH_DATA (push, src->depth);
PUSH_DATA (push, src->z);
} else {
src_ofst += src->y * src->pitch + src->x * cpp;
BEGIN_NVC0(push, NVC0_M2MF(PITCH_IN), 1);
PUSH_DATA (push, src->width * cpp);
exec |= NVC0_M2MF_EXEC_LINEAR_IN;
}
if (nouveau_bo_memtype(dst->bo)) {
BEGIN_NVC0(push, NVC0_M2MF(TILING_MODE_OUT), 5);
PUSH_DATA (push, dst->tile_mode);
PUSH_DATA (push, dst->width * cpp);
PUSH_DATA (push, dst->height);
PUSH_DATA (push, dst->depth);
PUSH_DATA (push, dst->z);
} else {
dst_ofst += dst->y * dst->pitch + dst->x * cpp;
BEGIN_NVC0(push, NVC0_M2MF(PITCH_OUT), 1);
PUSH_DATA (push, dst->width * cpp);
exec |= NVC0_M2MF_EXEC_LINEAR_OUT;
}
while (height) {
int line_count = height > 2047 ? 2047 : height;
BEGIN_NVC0(push, NVC0_M2MF(OFFSET_IN_HIGH), 2);
PUSH_DATAh(push, src->bo->offset + src_ofst);
PUSH_DATA (push, src->bo->offset + src_ofst);
BEGIN_NVC0(push, NVC0_M2MF(OFFSET_OUT_HIGH), 2);
PUSH_DATAh(push, dst->bo->offset + dst_ofst);
PUSH_DATA (push, dst->bo->offset + dst_ofst);
if (!(exec & NVC0_M2MF_EXEC_LINEAR_IN)) {
BEGIN_NVC0(push, NVC0_M2MF(TILING_POSITION_IN_X), 2);
PUSH_DATA (push, src->x * cpp);
PUSH_DATA (push, sy);
} else {
src_ofst += line_count * src->pitch;
}
if (!(exec & NVC0_M2MF_EXEC_LINEAR_OUT)) {
BEGIN_NVC0(push, NVC0_M2MF(TILING_POSITION_OUT_X), 2);
PUSH_DATA (push, dst->x * cpp);
PUSH_DATA (push, dy);
} else {
dst_ofst += line_count * dst->pitch;
}
BEGIN_NVC0(push, NVC0_M2MF(LINE_LENGTH_IN), 2);
PUSH_DATA (push, nblocksx * cpp);
PUSH_DATA (push, line_count);
BEGIN_NVC0(push, NVC0_M2MF(EXEC), 1);
PUSH_DATA (push, exec);
height -= line_count;
sy += line_count;
dy += line_count;
}
nouveau_bufctx_reset(bctx, 0);
}
static void
nve4_m2mf_transfer_rect(struct nvc0_context *nvc0,
const struct nv50_m2mf_rect *dst,
const struct nv50_m2mf_rect *src,
uint32_t nblocksx, uint32_t nblocksy)
{
struct nouveau_pushbuf *push = nvc0->base.pushbuf;
struct nouveau_bufctx *bctx = nvc0->bufctx;
uint32_t exec;
uint32_t src_base = src->base;
uint32_t dst_base = dst->base;
const int cpp = dst->cpp;
assert(dst->cpp == src->cpp);
nouveau_bufctx_refn(bctx, 0, dst->bo, dst->domain | NOUVEAU_BO_WR);
nouveau_bufctx_refn(bctx, 0, src->bo, src->domain | NOUVEAU_BO_RD);
nouveau_pushbuf_bufctx(push, bctx);
nouveau_pushbuf_validate(push);
exec = 0x200 /* 2D_ENABLE */ | 0x6 /* UNK */;
if (!nouveau_bo_memtype(dst->bo)) {
assert(!dst->z);
dst_base += dst->y * dst->pitch + dst->x * cpp;
exec |= 0x100; /* DST_MODE_2D_LINEAR */
}
if (!nouveau_bo_memtype(src->bo)) {
assert(!src->z);
src_base += src->y * src->pitch + src->x * cpp;
exec |= 0x080; /* SRC_MODE_2D_LINEAR */
}
BEGIN_NVC0(push, SUBC_COPY(0x070c), 6);
PUSH_DATA (push, 0x1000 | dst->tile_mode);
PUSH_DATA (push, dst->pitch);
PUSH_DATA (push, dst->height);
PUSH_DATA (push, dst->depth);
PUSH_DATA (push, dst->z);
PUSH_DATA (push, (dst->y << 16) | (dst->x * cpp));
BEGIN_NVC0(push, SUBC_COPY(0x0728), 6);
PUSH_DATA (push, 0x1000 | src->tile_mode);
PUSH_DATA (push, src->pitch);
PUSH_DATA (push, src->height);
PUSH_DATA (push, src->depth);
PUSH_DATA (push, src->z);
PUSH_DATA (push, (src->y << 16) | (src->x * cpp));
BEGIN_NVC0(push, SUBC_COPY(0x0400), 8);
PUSH_DATAh(push, src->bo->offset + src_base);
PUSH_DATA (push, src->bo->offset + src_base);
PUSH_DATAh(push, dst->bo->offset + dst_base);
PUSH_DATA (push, dst->bo->offset + dst_base);
PUSH_DATA (push, src->pitch);
PUSH_DATA (push, dst->pitch);
PUSH_DATA (push, nblocksx * cpp);
PUSH_DATA (push, nblocksy);
BEGIN_NVC0(push, SUBC_COPY(0x0300), 1);
PUSH_DATA (push, exec);
nouveau_bufctx_reset(bctx, 0);
}