/* Wait for write confirmation before writing data, but don't send
* an interrupt. */
- if (ctx->chip_class >= SI && data_sel != EOP_DATA_SEL_DISCARD)
+ if (data_sel != EOP_DATA_SEL_DISCARD)
sel |= EOP_INT_SEL(EOP_INT_SEL_SEND_DATA_AFTER_WR_CONFIRM);
if (ctx->chip_class >= GFX9) {
/* NOP waits for idle on Evergreen and later. */
if (rctx->chip_class >= CIK)
radeon_emit(cs, 0x00000000); /* NOP */
- else if (rctx->chip_class >= EVERGREEN)
+ else
radeon_emit(cs, 0xf0000000); /* NOP */
- else {
- /* TODO: R600-R700 should use the FENCE packet.
- * CS checker support is required. */
- }
}
void si_need_dma_space(struct r600_common_context *ctx, unsigned num_dw,
static const char *r600_get_family_name(const struct r600_common_screen *rscreen)
{
switch (rscreen->info.family) {
- case CHIP_R600: return "AMD R600";
- case CHIP_RV610: return "AMD RV610";
- case CHIP_RV630: return "AMD RV630";
- case CHIP_RV670: return "AMD RV670";
- case CHIP_RV620: return "AMD RV620";
- case CHIP_RV635: return "AMD RV635";
- case CHIP_RS780: return "AMD RS780";
- case CHIP_RS880: return "AMD RS880";
- case CHIP_RV770: return "AMD RV770";
- case CHIP_RV730: return "AMD RV730";
- case CHIP_RV710: return "AMD RV710";
- case CHIP_RV740: return "AMD RV740";
- case CHIP_CEDAR: return "AMD CEDAR";
- case CHIP_REDWOOD: return "AMD REDWOOD";
- case CHIP_JUNIPER: return "AMD JUNIPER";
- case CHIP_CYPRESS: return "AMD CYPRESS";
- case CHIP_HEMLOCK: return "AMD HEMLOCK";
- case CHIP_PALM: return "AMD PALM";
- case CHIP_SUMO: return "AMD SUMO";
- case CHIP_SUMO2: return "AMD SUMO2";
- case CHIP_BARTS: return "AMD BARTS";
- case CHIP_TURKS: return "AMD TURKS";
- case CHIP_CAICOS: return "AMD CAICOS";
- case CHIP_CAYMAN: return "AMD CAYMAN";
- case CHIP_ARUBA: return "AMD ARUBA";
case CHIP_TAHITI: return "AMD TAHITI";
case CHIP_PITCAIRN: return "AMD PITCAIRN";
case CHIP_VERDE: return "AMD CAPE VERDE";
const char *si_get_llvm_processor_name(enum radeon_family family)
{
switch (family) {
- case CHIP_R600:
- case CHIP_RV630:
- case CHIP_RV635:
- case CHIP_RV670:
- return "r600";
- case CHIP_RV610:
- case CHIP_RV620:
- case CHIP_RS780:
- case CHIP_RS880:
- return "rs880";
- case CHIP_RV710:
- return "rv710";
- case CHIP_RV730:
- return "rv730";
- case CHIP_RV740:
- case CHIP_RV770:
- return "rv770";
- case CHIP_PALM:
- case CHIP_CEDAR:
- return "cedar";
- case CHIP_SUMO:
- case CHIP_SUMO2:
- return "sumo";
- case CHIP_REDWOOD:
- return "redwood";
- case CHIP_JUNIPER:
- return "juniper";
- case CHIP_HEMLOCK:
- case CHIP_CYPRESS:
- return "cypress";
- case CHIP_BARTS:
- return "barts";
- case CHIP_TURKS:
- return "turks";
- case CHIP_CAICOS:
- return "caicos";
- case CHIP_CAYMAN:
- case CHIP_ARUBA:
- return "cayman";
-
case CHIP_TAHITI: return "tahiti";
case CHIP_PITCAIRN: return "pitcairn";
case CHIP_VERDE: return "verde";
case PIPE_COMPUTE_CAP_IR_TARGET: {
const char *gpu;
const char *triple;
- if (rscreen->family <= CHIP_ARUBA) {
- triple = "r600--";
- } else {
- if (HAVE_LLVM < 0x0400) {
- triple = "amdgcn--";
- } else {
- triple = "amdgcn-mesa-mesa3d";
- }
- }
- switch(rscreen->family) {
- /* Clang < 3.6 is missing Hainan in its list of
- * GPUs, so we need to use the name of a similar GPU.
- */
- default:
- gpu = si_get_llvm_processor_name(rscreen->family);
- break;
- }
+
+ if (HAVE_LLVM < 0x0400)
+ triple = "amdgcn--";
+ else
+ triple = "amdgcn-mesa-mesa3d";
+
+ gpu = si_get_llvm_processor_name(rscreen->family);
if (ret) {
sprintf(ret, "%s-%s", gpu, triple);
}
case PIPE_COMPUTE_CAP_SUBGROUP_SIZE:
if (ret) {
uint32_t *subgroup_size = ret;
- *subgroup_size = r600_wavefront_size(rscreen->family);
+ *subgroup_size = 64;
}
return sizeof(uint32_t);
case PIPE_COMPUTE_CAP_MAX_VARIABLE_THREADS_PER_BLOCK:
return 4;
}
-static inline unsigned r600_wavefront_size(enum radeon_family family)
-{
- switch (family) {
- case CHIP_RV610:
- case CHIP_RS780:
- case CHIP_RV620:
- case CHIP_RS880:
- return 16;
- case CHIP_RV630:
- case CHIP_RV635:
- case CHIP_RV730:
- case CHIP_RV710:
- case CHIP_PALM:
- case CHIP_CEDAR:
- return 32;
- default:
- return 64;
- }
-}
-
static inline enum radeon_bo_priority
r600_get_sampler_view_priority(struct r600_resource *res)
{
emit_sample_streamout(cs, va + 32 * stream, stream);
break;
case PIPE_QUERY_TIME_ELAPSED:
- if (ctx->chip_class >= SI) {
- /* Write the timestamp from the CP not waiting for
- * outstanding draws (top-of-pipe).
- */
- radeon_emit(cs, PKT3(PKT3_COPY_DATA, 4, 0));
- radeon_emit(cs, COPY_DATA_COUNT_SEL |
- COPY_DATA_SRC_SEL(COPY_DATA_TIMESTAMP) |
- COPY_DATA_DST_SEL(COPY_DATA_MEM_ASYNC));
- radeon_emit(cs, 0);
- radeon_emit(cs, 0);
- radeon_emit(cs, va);
- radeon_emit(cs, va >> 32);
- } else {
- /* Write the timestamp after the last draw is done.
- * (bottom-of-pipe)
- */
- si_gfx_write_event_eop(ctx, EVENT_TYPE_BOTTOM_OF_PIPE_TS,
- 0, EOP_DATA_SEL_TIMESTAMP,
- NULL, va, 0, query->b.type);
- }
+ /* Write the timestamp from the CP not waiting for
+ * outstanding draws (top-of-pipe).
+ */
+ radeon_emit(cs, PKT3(PKT3_COPY_DATA, 4, 0));
+ radeon_emit(cs, COPY_DATA_COUNT_SEL |
+ COPY_DATA_SRC_SEL(COPY_DATA_TIMESTAMP) |
+ COPY_DATA_DST_SEL(COPY_DATA_MEM_ASYNC));
+ radeon_emit(cs, 0);
+ radeon_emit(cs, 0);
+ radeon_emit(cs, va);
+ radeon_emit(cs, va >> 32);
break;
case PIPE_QUERY_PIPELINE_STATISTICS:
radeon_emit(cs, PKT3(PKT3_EVENT_WRITE, 2, 0));
num_bufs * 11; /* STRMOUT_BUFFER_UPDATE, BUFFER_SIZE */
begin->num_dw = 12; /* flush_vgt_streamout */
-
- if (rctx->chip_class >= SI) {
- begin->num_dw += num_bufs * 4; /* SET_CONTEXT_REG */
- } else {
- begin->num_dw += num_bufs * 7; /* SET_CONTEXT_REG */
-
- if (rctx->family >= CHIP_RS780 && rctx->family <= CHIP_RV740)
- begin->num_dw += num_bufs * 5; /* STRMOUT_BASE_UPDATE */
- }
-
+ begin->num_dw += num_bufs * 4; /* SET_CONTEXT_REG */
begin->num_dw +=
num_bufs_appended * 8 + /* STRMOUT_BUFFER_UPDATE */
- (num_bufs - num_bufs_appended) * 6 + /* STRMOUT_BUFFER_UPDATE */
- (rctx->family > CHIP_R600 && rctx->family < CHIP_RS780 ? 2 : 0); /* SURFACE_BASE_UPDATE */
+ (num_bufs - num_bufs_appended) * 6; + /* STRMOUT_BUFFER_UPDATE */
rctx->set_atom_dirty(rctx, begin, true);
/* The register is at different places on different ASICs. */
if (rctx->chip_class >= CIK) {
reg_strmout_cntl = R_0300FC_CP_STRMOUT_CNTL;
- } else if (rctx->chip_class >= EVERGREEN) {
- reg_strmout_cntl = R_0084FC_CP_STRMOUT_CNTL;
- } else {
- reg_strmout_cntl = R_008490_CP_STRMOUT_CNTL;
- }
-
- if (rctx->chip_class >= CIK) {
radeon_set_uconfig_reg(cs, reg_strmout_cntl, 0);
} else {
+ reg_strmout_cntl = R_0084FC_CP_STRMOUT_CNTL;
radeon_set_config_reg(cs, reg_strmout_cntl, 0);
}
struct radeon_winsys_cs *cs = rctx->gfx.cs;
struct r600_so_target **t = rctx->streamout.targets;
uint16_t *stride_in_dw = rctx->streamout.stride_in_dw;
- unsigned i, update_flags = 0;
+ unsigned i;
r600_flush_vgt_streamout(rctx);
t[i]->stride_in_dw = stride_in_dw[i];
- if (rctx->chip_class >= SI) {
- /* SI binds streamout buffers as shader resources.
- * VGT only counts primitives and tells the shader
- * through SGPRs what to do. */
- radeon_set_context_reg_seq(cs, R_028AD0_VGT_STRMOUT_BUFFER_SIZE_0 + 16*i, 2);
- radeon_emit(cs, (t[i]->b.buffer_offset +
- t[i]->b.buffer_size) >> 2); /* BUFFER_SIZE (in DW) */
- radeon_emit(cs, stride_in_dw[i]); /* VTX_STRIDE (in DW) */
- } else {
- uint64_t va = r600_resource(t[i]->b.buffer)->gpu_address;
-
- update_flags |= SURFACE_BASE_UPDATE_STRMOUT(i);
-
- radeon_set_context_reg_seq(cs, R_028AD0_VGT_STRMOUT_BUFFER_SIZE_0 + 16*i, 3);
- radeon_emit(cs, (t[i]->b.buffer_offset +
- t[i]->b.buffer_size) >> 2); /* BUFFER_SIZE (in DW) */
- radeon_emit(cs, stride_in_dw[i]); /* VTX_STRIDE (in DW) */
- radeon_emit(cs, va >> 8); /* BUFFER_BASE */
-
- r600_emit_reloc(rctx, &rctx->gfx, r600_resource(t[i]->b.buffer),
- RADEON_USAGE_WRITE, RADEON_PRIO_SHADER_RW_BUFFER);
-
- /* R7xx requires this packet after updating BUFFER_BASE.
- * Without this, R7xx locks up. */
- if (rctx->family >= CHIP_RS780 && rctx->family <= CHIP_RV740) {
- radeon_emit(cs, PKT3(PKT3_STRMOUT_BASE_UPDATE, 1, 0));
- radeon_emit(cs, i);
- radeon_emit(cs, va >> 8);
-
- r600_emit_reloc(rctx, &rctx->gfx, r600_resource(t[i]->b.buffer),
- RADEON_USAGE_WRITE, RADEON_PRIO_SHADER_RW_BUFFER);
- }
- }
+ /* SI binds streamout buffers as shader resources.
+ * VGT only counts primitives and tells the shader
+ * through SGPRs what to do. */
+ radeon_set_context_reg_seq(cs, R_028AD0_VGT_STRMOUT_BUFFER_SIZE_0 + 16*i, 2);
+ radeon_emit(cs, (t[i]->b.buffer_offset +
+ t[i]->b.buffer_size) >> 2); /* BUFFER_SIZE (in DW) */
+ radeon_emit(cs, stride_in_dw[i]); /* VTX_STRIDE (in DW) */
if (rctx->streamout.append_bitmask & (1 << i) && t[i]->buf_filled_size_valid) {
uint64_t va = t[i]->buf_filled_size->gpu_address +
}
}
- if (rctx->family > CHIP_R600 && rctx->family < CHIP_RV770) {
- radeon_emit(cs, PKT3(PKT3_SURFACE_BASE_UPDATE, 0, 0));
- radeon_emit(cs, update_flags);
- }
rctx->streamout.begin_emitted = true;
}
static void r600_emit_streamout_enable(struct r600_common_context *rctx,
struct r600_atom *atom)
{
- unsigned strmout_config_reg = R_028AB0_VGT_STRMOUT_EN;
- unsigned strmout_config_val = S_028B94_STREAMOUT_0_EN(r600_get_strmout_en(rctx));
- unsigned strmout_buffer_reg = R_028B20_VGT_STRMOUT_BUFFER_EN;
- unsigned strmout_buffer_val = rctx->streamout.hw_enabled_mask &
- rctx->streamout.enabled_stream_buffers_mask;
-
- if (rctx->chip_class >= EVERGREEN) {
- strmout_buffer_reg = R_028B98_VGT_STRMOUT_BUFFER_CONFIG;
-
- strmout_config_reg = R_028B94_VGT_STRMOUT_CONFIG;
- strmout_config_val |=
- S_028B94_RAST_STREAM(0) |
- S_028B94_STREAMOUT_1_EN(r600_get_strmout_en(rctx)) |
- S_028B94_STREAMOUT_2_EN(r600_get_strmout_en(rctx)) |
- S_028B94_STREAMOUT_3_EN(r600_get_strmout_en(rctx));
- }
- radeon_set_context_reg(rctx->gfx.cs, strmout_buffer_reg, strmout_buffer_val);
- radeon_set_context_reg(rctx->gfx.cs, strmout_config_reg, strmout_config_val);
+ radeon_set_context_reg_seq(rctx->gfx.cs, R_028B94_VGT_STRMOUT_CONFIG, 2);
+ radeon_emit(rctx->gfx.cs,
+ S_028B94_STREAMOUT_0_EN(r600_get_strmout_en(rctx)) |
+ S_028B94_RAST_STREAM(0) |
+ S_028B94_STREAMOUT_1_EN(r600_get_strmout_en(rctx)) |
+ S_028B94_STREAMOUT_2_EN(r600_get_strmout_en(rctx)) |
+ S_028B94_STREAMOUT_3_EN(r600_get_strmout_en(rctx)));
+ radeon_emit(rctx->gfx.cs,
+ rctx->streamout.hw_enabled_mask &
+ rctx->streamout.enabled_stream_buffers_mask);
}
static void r600_set_streamout_enable(struct r600_common_context *rctx, bool enable)
templ.bind |= new_bind_flag;
- /* r600g doesn't react to dirty_tex_descriptor_counter */
- if (rctx->chip_class < SI)
- return;
-
if (rtex->resource.b.is_shared)
return;
switch(u_reduce_video_profile(templ->profile)) {
case PIPE_VIDEO_FORMAT_MPEG12:
- if (templ->entrypoint > PIPE_VIDEO_ENTRYPOINT_BITSTREAM || info.family < CHIP_PALM)
+ if (templ->entrypoint > PIPE_VIDEO_ENTRYPOINT_BITSTREAM)
return vl_create_mpeg12_decoder(context, templ);
/* fall through */
case PIPE_VIDEO_FORMAT_MPEG12:
return profile != PIPE_VIDEO_PROFILE_MPEG1;
case PIPE_VIDEO_FORMAT_MPEG4:
- /* no support for MPEG4 on older hw */
- return rscreen->family >= CHIP_PALM;
+ return 1;
case PIPE_VIDEO_FORMAT_MPEG4_AVC:
if ((rscreen->family == CHIP_POLARIS10 ||
rscreen->family == CHIP_POLARIS11) &&
return PIPE_FORMAT_NV12;
case PIPE_VIDEO_CAP_PREFERS_INTERLACED:
- case PIPE_VIDEO_CAP_SUPPORTS_INTERLACED:
- if (rscreen->family < CHIP_PALM) {
- /* MPEG2 only with shaders and no support for
- interlacing on R6xx style UVD */
- return codec != PIPE_VIDEO_FORMAT_MPEG12 &&
- rscreen->family > CHIP_RV770;
- } else {
- enum pipe_video_format format = u_reduce_video_profile(profile);
-
- if (format == PIPE_VIDEO_FORMAT_HEVC)
- return false; //The firmware doesn't support interlaced HEVC.
- else if (format == PIPE_VIDEO_FORMAT_JPEG)
- return false;
- return true;
- }
+ case PIPE_VIDEO_CAP_SUPPORTS_INTERLACED: {
+ enum pipe_video_format format = u_reduce_video_profile(profile);
+
+ if (format == PIPE_VIDEO_FORMAT_HEVC)
+ return false; //The firmware doesn't support interlaced HEVC.
+ else if (format == PIPE_VIDEO_FORMAT_JPEG)
+ return false;
+ return true;
+ }
case PIPE_VIDEO_CAP_SUPPORTS_PROGRESSIVE:
return true;
case PIPE_VIDEO_CAP_MAX_LEVEL: