From: Marek Olšák <[email protected]>
v2: fix enabling primitive binning
---
src/amd/addrlib/amdgpu_asic_addr.h | 2 ++
src/amd/addrlib/gfx9/gfx9addrlib.cpp | 2 +-
src/amd/common/ac_gpu_info.c | 6 ++++++
src/amd/common/ac_llvm_util.c | 2 ++
src/amd/common/ac_surface.c | 4 ++++
src/amd/common/amd_family.h | 1 +
src/amd/common/gfx9d.h | 3 +++
src/gallium/drivers/radeonsi/si_pipe.c | 10 ++++++----
src/gallium/drivers/radeonsi/si_state.c | 4 +++-
src/gallium/drivers/radeonsi/si_state_binning.c | 1 +
10 files changed, 29 insertions(+), 6 deletions(-)
diff --git a/src/amd/addrlib/amdgpu_asic_addr.h
b/src/amd/addrlib/amdgpu_asic_addr.h
index e5838d42a3c..7436c5493e1 100644
--- a/src/amd/addrlib/amdgpu_asic_addr.h
+++ b/src/amd/addrlib/amdgpu_asic_addr.h
@@ -83,20 +83,21 @@
#define AMDGPU_CARRIZO_RANGE 0x01, 0x21
#define AMDGPU_BRISTOL_RANGE 0x10, 0x21
#define AMDGPU_STONEY_RANGE 0x61, 0xFF
#define AMDGPU_VEGA10_RANGE 0x01, 0x14
#define AMDGPU_VEGA12_RANGE 0x14, 0x28
#define AMDGPU_VEGA20_RANGE 0x28, 0xFF
#define AMDGPU_RAVEN_RANGE 0x01, 0x81
+#define AMDGPU_RAVEN2_RANGE 0x81, 0xFF
#define AMDGPU_EXPAND_FIX(x) x
#define AMDGPU_RANGE_HELPER(val, min, max) ((val >= min) && (val < max))
#define AMDGPU_IN_RANGE(val, ...)
AMDGPU_EXPAND_FIX(AMDGPU_RANGE_HELPER(val, __VA_ARGS__))
// ASICREV_IS(eRevisionId, revisionName)
#define ASICREV_IS(r, rn) AMDGPU_IN_RANGE(r, AMDGPU_##rn##_RANGE)
#define ASICREV_IS_TAHITI_P(r) ASICREV_IS(r, TAHITI)
#define ASICREV_IS_PITCAIRN_PM(r) ASICREV_IS(r, PITCAIRN)
@@ -125,12 +126,13 @@
#define ASICREV_IS_CARRIZO_BRISTOL(r) ASICREV_IS(r, BRISTOL)
#define ASICREV_IS_STONEY(r) ASICREV_IS(r, STONEY)
#define ASICREV_IS_VEGA10_M(r) ASICREV_IS(r, VEGA10)
#define ASICREV_IS_VEGA10_P(r) ASICREV_IS(r, VEGA10)
#define ASICREV_IS_VEGA12_P(r) ASICREV_IS(r, VEGA12)
#define ASICREV_IS_VEGA12_p(r) ASICREV_IS(r, VEGA12)
#define ASICREV_IS_VEGA20_P(r) ASICREV_IS(r, VEGA20)
#define ASICREV_IS_RAVEN(r) ASICREV_IS(r, RAVEN)
+#define ASICREV_IS_RAVEN2(r) ASICREV_IS(r, RAVEN2)
#endif // _AMDGPU_ASIC_ADDR_H
diff --git a/src/amd/addrlib/gfx9/gfx9addrlib.cpp
b/src/amd/addrlib/gfx9/gfx9addrlib.cpp
index d27aabbb60c..f115242c89c 100644
--- a/src/amd/addrlib/gfx9/gfx9addrlib.cpp
+++ b/src/amd/addrlib/gfx9/gfx9addrlib.cpp
@@ -1284,21 +1284,21 @@ ChipFamily Gfx9Lib::HwlConvertChipFamily(
m_settings.htileAlignFix = 1;
m_settings.applyAliasFix = 1;
}
m_settings.metaBaseAlignFix = 1;
m_settings.depthPipeXorDisable = 1;
break;
case FAMILY_RV:
m_settings.isArcticIsland = 1;
- m_settings.isRaven = ASICREV_IS_RAVEN(uChipRevision);
+ m_settings.isRaven = ASICREV_IS_RAVEN(uChipRevision) ||
ASICREV_IS_RAVEN2(uChipRevision);
if (m_settings.isRaven)
{
m_settings.isDcn1 = 1;
}
m_settings.metaBaseAlignFix = 1;
if (ASICREV_IS_RAVEN(uChipRevision))
{
diff --git a/src/amd/common/ac_gpu_info.c b/src/amd/common/ac_gpu_info.c
index 2c70fb2c721..689f544c18b 100644
--- a/src/amd/common/ac_gpu_info.c
+++ b/src/amd/common/ac_gpu_info.c
@@ -307,20 +307,26 @@ bool ac_query_gpu_info(int fd, amdgpu_device_handle dev,
info->name = #cfamily; \
break;
#include "pci_ids/radeonsi_pci_ids.h"
#undef CHIPSET
default:
fprintf(stderr, "amdgpu: Invalid PCI ID.\n");
return false;
}
+ /* Raven2 uses the same PCI IDs as Raven1, but different revision IDs. */
+ if (info->family == CHIP_RAVEN && amdinfo->chip_rev >= 0x8) {
+ info->family = CHIP_RAVEN2;
+ info->name = "RAVEN2";
+ }
+
if (info->family >= CHIP_VEGA10)
info->chip_class = GFX9;
else if (info->family >= CHIP_TONGA)
info->chip_class = VI;
else if (info->family >= CHIP_BONAIRE)
info->chip_class = CIK;
else if (info->family >= CHIP_TAHITI)
info->chip_class = SI;
else {
fprintf(stderr, "amdgpu: Unknown family.\n");
diff --git a/src/amd/common/ac_llvm_util.c b/src/amd/common/ac_llvm_util.c
index cd3525187a0..69d9f7b9f3f 100644
--- a/src/amd/common/ac_llvm_util.c
+++ b/src/amd/common/ac_llvm_util.c
@@ -128,20 +128,22 @@ const char *ac_get_llvm_processor_name(enum radeon_family
family)
case CHIP_VEGAM:
return "polaris11";
case CHIP_VEGA10:
return "gfx900";
case CHIP_RAVEN:
return "gfx902";
case CHIP_VEGA12:
return HAVE_LLVM >= 0x0700 ? "gfx904" : "gfx902";
case CHIP_VEGA20:
return HAVE_LLVM >= 0x0700 ? "gfx906" : "gfx902";
+ case CHIP_RAVEN2:
+ return "gfx902"; /* TODO: use gfx909 when it's available */
default:
return "";
}
}
static LLVMTargetMachineRef ac_create_target_machine(enum radeon_family family,
enum
ac_target_machine_options tm_options,
LLVMCodeGenOptLevel level,
const char **out_triple)
{
diff --git a/src/amd/common/ac_surface.c b/src/amd/common/ac_surface.c
index 94723dc9c09..1f7e2344625 100644
--- a/src/amd/common/ac_surface.c
+++ b/src/amd/common/ac_surface.c
@@ -144,20 +144,24 @@ static void addrlib_family_rev_id(enum radeon_family
family,
*addrlib_revid = get_first(AMDGPU_VEGA12_RANGE);
break;
case CHIP_VEGA20:
*addrlib_family = FAMILY_AI;
*addrlib_revid = get_first(AMDGPU_VEGA20_RANGE);
break;
case CHIP_RAVEN:
*addrlib_family = FAMILY_RV;
*addrlib_revid = get_first(AMDGPU_RAVEN_RANGE);
break;
+ case CHIP_RAVEN2:
+ *addrlib_family = FAMILY_RV;
+ *addrlib_revid = get_first(AMDGPU_RAVEN2_RANGE);
+ break;
default:
fprintf(stderr, "amdgpu: Unknown family.\n");
}
}
static void *ADDR_API allocSysMem(const ADDR_ALLOCSYSMEM_INPUT * pInput)
{
return malloc(pInput->sizeInBytes);
}
diff --git a/src/amd/common/amd_family.h b/src/amd/common/amd_family.h
index a282898be06..185ba029763 100644
--- a/src/amd/common/amd_family.h
+++ b/src/amd/common/amd_family.h
@@ -90,20 +90,21 @@ enum radeon_family {
CHIP_FIJI,
CHIP_STONEY,
CHIP_POLARIS10,
CHIP_POLARIS11,
CHIP_POLARIS12,
CHIP_VEGAM,
CHIP_VEGA10,
CHIP_VEGA12,
CHIP_VEGA20,
CHIP_RAVEN,
+ CHIP_RAVEN2,
CHIP_LAST,
};
enum chip_class {
CLASS_UNKNOWN = 0,
R300,
R400,
R500,
R600,
R700,
diff --git a/src/amd/common/gfx9d.h b/src/amd/common/gfx9d.h
index d18e6655d33..2e790c54699 100644
--- a/src/amd/common/gfx9d.h
+++ b/src/amd/common/gfx9d.h
@@ -4450,20 +4450,23 @@
#define R_028424_CB_DCC_CONTROL
0x028424
#define S_028424_OVERWRITE_COMBINER_DISABLE(x) (((unsigned)(x)
& 0x1) << 0)
#define G_028424_OVERWRITE_COMBINER_DISABLE(x) (((x) >> 0)
& 0x1)
#define C_028424_OVERWRITE_COMBINER_DISABLE
0xFFFFFFFE
#define S_028424_OVERWRITE_COMBINER_MRT_SHARING_DISABLE(x) (((unsigned)(x)
& 0x1) << 1)
#define G_028424_OVERWRITE_COMBINER_MRT_SHARING_DISABLE(x) (((x) >> 1)
& 0x1)
#define C_028424_OVERWRITE_COMBINER_MRT_SHARING_DISABLE
0xFFFFFFFD
#define S_028424_OVERWRITE_COMBINER_WATERMARK(x) (((unsigned)(x)
& 0x1F) << 2)
#define G_028424_OVERWRITE_COMBINER_WATERMARK(x) (((x) >> 2)
& 0x1F)
#define C_028424_OVERWRITE_COMBINER_WATERMARK
0xFFFFFF83
+#define S_028424_DISABLE_CONSTANT_ENCODE_REG(x) (((unsigned)(x)
& 0x1) << 10) /* Raven2+ */
+#define G_028424_DISABLE_CONSTANT_ENCODE_REG(x) (((x) >> 10)
& 0x1)
+#define C_028424_DISABLE_CONSTANT_ENCODE_REG
0xFFFFFBFF
#define R_02842C_DB_STENCIL_CONTROL
0x02842C
#define S_02842C_STENCILFAIL(x) (((unsigned)(x)
& 0x0F) << 0)
#define G_02842C_STENCILFAIL(x) (((x) >> 0)
& 0x0F)
#define C_02842C_STENCILFAIL
0xFFFFFFF0
#define S_02842C_STENCILZPASS(x) (((unsigned)(x)
& 0x0F) << 4)
#define G_02842C_STENCILZPASS(x) (((x) >> 4)
& 0x0F)
#define C_02842C_STENCILZPASS
0xFFFFFF0F
#define S_02842C_STENCILZFAIL(x) (((unsigned)(x)
& 0x0F) << 8)
#define G_02842C_STENCILZFAIL(x) (((x) >> 8)
& 0x0F)
#define C_02842C_STENCILZFAIL
0xFFFFF0FF
diff --git a/src/gallium/drivers/radeonsi/si_pipe.c
b/src/gallium/drivers/radeonsi/si_pipe.c
index 6118b8076f1..490a3714836 100644
--- a/src/gallium/drivers/radeonsi/si_pipe.c
+++ b/src/gallium/drivers/radeonsi/si_pipe.c
@@ -1026,24 +1026,25 @@ struct pipe_screen *radeonsi_screen_create(struct
radeon_winsys *ws,
sscreen->has_msaa_sample_loc_bug = (sscreen->info.family >= CHIP_POLARIS10
&&
sscreen->info.family <=
CHIP_POLARIS12) ||
sscreen->info.family == CHIP_VEGA10
||
sscreen->info.family == CHIP_RAVEN;
sscreen->has_ls_vgpr_init_bug = sscreen->info.family == CHIP_VEGA10 ||
sscreen->info.family == CHIP_RAVEN;
if (sscreen->debug_flags & DBG(DPBB)) {
sscreen->dpbb_allowed = true;
} else {
- /* Only enable primitive binning on Raven by default. */
+ /* Only enable primitive binning on APUs by default. */
/* TODO: Investigate if binning is profitable on Vega12. */
- sscreen->dpbb_allowed = sscreen->info.family == CHIP_RAVEN &&
- !(sscreen->debug_flags & DBG(NO_DPBB));
+ sscreen->dpbb_allowed = !(sscreen->debug_flags & DBG(NO_DPBB))
&&
+ (sscreen->info.family == CHIP_RAVEN ||
+ sscreen->info.family == CHIP_RAVEN2);
}
if (sscreen->debug_flags & DBG(DFSM)) {
sscreen->dfsm_allowed = sscreen->dpbb_allowed;
} else {
sscreen->dfsm_allowed = sscreen->dpbb_allowed &&
!(sscreen->debug_flags & DBG(NO_DFSM));
}
/* While it would be nice not to have this flag, we are constrained
@@ -1056,21 +1057,22 @@ struct pipe_screen *radeonsi_screen_create(struct
radeon_winsys *ws,
* always disable it.
*/
if (sscreen->info.family == CHIP_STONEY ||
sscreen->info.chip_class >= GFX9) {
sscreen->has_rbplus = true;
sscreen->rbplus_allowed =
!(sscreen->debug_flags & DBG(NO_RB_PLUS)) &&
(sscreen->info.family == CHIP_STONEY ||
sscreen->info.family == CHIP_VEGA12 ||
- sscreen->info.family == CHIP_RAVEN);
+ sscreen->info.family == CHIP_RAVEN ||
+ sscreen->info.family == CHIP_RAVEN2);
}
sscreen->dcc_msaa_allowed =
!(sscreen->debug_flags & DBG(NO_DCC_MSAA));
sscreen->cpdma_prefetch_writes_memory = sscreen->info.chip_class <= VI;
(void) mtx_init(&sscreen->shader_parts_mutex, mtx_plain);
sscreen->use_monolithic_shaders =
(sscreen->debug_flags & DBG(MONOLITHIC_SHADERS)) != 0;
diff --git a/src/gallium/drivers/radeonsi/si_state.c
b/src/gallium/drivers/radeonsi/si_state.c
index 43d76d19916..0293bdfa791 100644
--- a/src/gallium/drivers/radeonsi/si_state.c
+++ b/src/gallium/drivers/radeonsi/si_state.c
@@ -113,21 +113,22 @@ static void si_emit_cb_render_state(struct si_context
*sctx)
blend &&
blend->blend_enable_4bit & cb_target_mask &&
sctx->framebuffer.nr_samples >= 2;
unsigned watermark =
sctx->framebuffer.dcc_overwrite_combiner_watermark;
radeon_opt_set_context_reg(
sctx, R_028424_CB_DCC_CONTROL,
SI_TRACKED_CB_DCC_CONTROL,
S_028424_OVERWRITE_COMBINER_MRT_SHARING_DISABLE(1) |
S_028424_OVERWRITE_COMBINER_WATERMARK(watermark) |
-
S_028424_OVERWRITE_COMBINER_DISABLE(oc_disable));
+ S_028424_OVERWRITE_COMBINER_DISABLE(oc_disable)
|
+
S_028424_DISABLE_CONSTANT_ENCODE_REG(sctx->family == CHIP_RAVEN2));
}
/* RB+ register settings. */
if (sctx->screen->rbplus_allowed) {
unsigned spi_shader_col_format =
sctx->ps_shader.cso ?
sctx->ps_shader.current->key.part.ps.epilog.spi_shader_col_format : 0;
unsigned sx_ps_downconvert = 0;
unsigned sx_blend_opt_epsilon = 0;
unsigned sx_blend_opt_control = 0;
@@ -5093,20 +5094,21 @@ static void si_init_config(struct si_context *sctx)
unsigned num_se = sscreen->info.max_se;
unsigned pc_lines = 0;
switch (sctx->family) {
case CHIP_VEGA10:
case CHIP_VEGA12:
case CHIP_VEGA20:
pc_lines = 4096;
break;
case CHIP_RAVEN:
+ case CHIP_RAVEN2:
pc_lines = 1024;
break;
default:
assert(0);
}
si_pm4_set_reg(pm4, R_028C48_PA_SC_BINNER_CNTL_1,
S_028C48_MAX_ALLOC_COUNT(MIN2(128, pc_lines / (4
* num_se))) |
S_028C48_MAX_PRIM_PER_BATCH(1023));
si_pm4_set_reg(pm4,
R_028C4C_PA_SC_CONSERVATIVE_RASTERIZATION_CNTL,
diff --git a/src/gallium/drivers/radeonsi/si_state_binning.c
b/src/gallium/drivers/radeonsi/si_state_binning.c
index 70c129242d1..3516e561282 100644
--- a/src/gallium/drivers/radeonsi/si_state_binning.c
+++ b/src/gallium/drivers/radeonsi/si_state_binning.c
@@ -400,20 +400,21 @@ void si_emit_dpbb_state(struct si_context *sctx)
/* Tunable parameters. Also test with DFSM enabled/disabled. */
unsigned context_states_per_bin; /* allowed range: [0, 5] */
unsigned persistent_states_per_bin; /* allowed range: [0, 31] */
unsigned fpovs_per_batch; /* allowed range: [0, 255], 0 = unlimited */
switch (sctx->family) {
case CHIP_VEGA10:
case CHIP_VEGA12:
case CHIP_VEGA20:
case CHIP_RAVEN:
+ case CHIP_RAVEN2:
/* Tuned for Raven. Vega might need different values. */
context_states_per_bin = 5;
persistent_states_per_bin = 31;
fpovs_per_batch = 63;
break;
default:
assert(0);
}
/* Emit registers. */
