summaryrefslogtreecommitdiffstats
path: root/src/intel
diff options
context:
space:
mode:
Diffstat (limited to 'src/intel')
-rw-r--r--src/intel/compiler/brw_eu.c29
-rw-r--r--src/intel/compiler/brw_eu.h3
-rw-r--r--src/intel/compiler/brw_eu_emit.c98
-rw-r--r--src/intel/compiler/brw_fs_generator.cpp4
4 files changed, 79 insertions, 55 deletions
diff --git a/src/intel/compiler/brw_eu.c b/src/intel/compiler/brw_eu.c
index 3646076a8e8..5375209d4fc 100644
--- a/src/intel/compiler/brw_eu.c
+++ b/src/intel/compiler/brw_eu.c
@@ -126,6 +126,35 @@ brw_swizzle_immediate(enum brw_reg_type type, uint32_t x, unsigned swz)
}
}
+unsigned
+brw_get_default_exec_size(struct brw_codegen *p)
+{
+ return brw_inst_exec_size(p->devinfo, p->current);
+}
+
+unsigned
+brw_get_default_group(struct brw_codegen *p)
+{
+ if (p->devinfo->gen >= 6) {
+ unsigned group = brw_inst_qtr_control(p->devinfo, p->current) * 8;
+ if (p->devinfo->gen >= 7)
+ group += brw_inst_nib_control(p->devinfo, p->current) * 4;
+ return group;
+ } else {
+ unsigned qtr_control = brw_inst_qtr_control(p->devinfo, p->current);
+ if (qtr_control == BRW_COMPRESSION_COMPRESSED)
+ return 0;
+ else
+ return qtr_control * 8;
+ }
+}
+
+unsigned
+brw_get_default_access_mode(struct brw_codegen *p)
+{
+ return brw_inst_access_mode(p->devinfo, p->current);
+}
+
void
brw_set_default_exec_size(struct brw_codegen *p, unsigned value)
{
diff --git a/src/intel/compiler/brw_eu.h b/src/intel/compiler/brw_eu.h
index 84d5a6f86be..d709d12b671 100644
--- a/src/intel/compiler/brw_eu.h
+++ b/src/intel/compiler/brw_eu.h
@@ -106,6 +106,9 @@ struct brw_codegen {
void brw_pop_insn_state( struct brw_codegen *p );
void brw_push_insn_state( struct brw_codegen *p );
+unsigned brw_get_default_exec_size(struct brw_codegen *p);
+unsigned brw_get_default_group(struct brw_codegen *p);
+unsigned brw_get_default_access_mode(struct brw_codegen *p);
void brw_set_default_exec_size(struct brw_codegen *p, unsigned value);
void brw_set_default_mask_control( struct brw_codegen *p, unsigned value );
void brw_set_default_saturate( struct brw_codegen *p, bool enable );
diff --git a/src/intel/compiler/brw_eu_emit.c b/src/intel/compiler/brw_eu_emit.c
index c442f8cc6f4..80eea9ecba5 100644
--- a/src/intel/compiler/brw_eu_emit.c
+++ b/src/intel/compiler/brw_eu_emit.c
@@ -997,7 +997,7 @@ brw_MOV(struct brw_codegen *p, struct brw_reg dest, struct brw_reg src0)
* each element twice.
*/
if (devinfo->gen == 7 && !devinfo->is_haswell &&
- brw_inst_access_mode(devinfo, p->current) == BRW_ALIGN_1 &&
+ brw_get_default_access_mode(p) == BRW_ALIGN_1 &&
dest.type == BRW_REGISTER_TYPE_DF &&
(src0.type == BRW_REGISTER_TYPE_F ||
src0.type == BRW_REGISTER_TYPE_D ||
@@ -1119,7 +1119,7 @@ brw_inst *
brw_F32TO16(struct brw_codegen *p, struct brw_reg dst, struct brw_reg src)
{
const struct gen_device_info *devinfo = p->devinfo;
- const bool align16 = brw_inst_access_mode(devinfo, p->current) == BRW_ALIGN_16;
+ const bool align16 = brw_get_default_access_mode(p) == BRW_ALIGN_16;
/* The F32TO16 instruction doesn't support 32-bit destination types in
* Align1 mode, and neither does the Gen8 implementation in terms of a
* converting MOV. Gen7 does zero out the high 16 bits in Align16 mode as
@@ -1166,7 +1166,7 @@ brw_inst *
brw_F16TO32(struct brw_codegen *p, struct brw_reg dst, struct brw_reg src)
{
const struct gen_device_info *devinfo = p->devinfo;
- bool align16 = brw_inst_access_mode(devinfo, p->current) == BRW_ALIGN_16;
+ bool align16 = brw_get_default_access_mode(p) == BRW_ALIGN_16;
if (align16) {
assert(src.type == BRW_REGISTER_TYPE_UD);
@@ -1337,8 +1337,7 @@ gen6_IF(struct brw_codegen *p, enum brw_conditional_mod conditional,
insn = next_insn(p, BRW_OPCODE_IF);
brw_set_dest(p, insn, brw_imm_w(0));
- brw_inst_set_exec_size(devinfo, insn,
- brw_inst_exec_size(devinfo, p->current));
+ brw_inst_set_exec_size(devinfo, insn, brw_get_default_exec_size(p));
brw_inst_set_gen6_jump_count(devinfo, insn, 0);
brw_set_src0(p, insn, src0);
brw_set_src1(p, insn, src1);
@@ -1624,8 +1623,7 @@ brw_BREAK(struct brw_codegen *p)
p->if_depth_in_loop[p->loop_stack_depth]);
}
brw_inst_set_qtr_control(devinfo, insn, BRW_COMPRESSION_NONE);
- brw_inst_set_exec_size(devinfo, insn,
- brw_inst_exec_size(devinfo, p->current));
+ brw_inst_set_exec_size(devinfo, insn, brw_get_default_exec_size(p));
return insn;
}
@@ -1650,8 +1648,7 @@ brw_CONT(struct brw_codegen *p)
p->if_depth_in_loop[p->loop_stack_depth]);
}
brw_inst_set_qtr_control(devinfo, insn, BRW_COMPRESSION_NONE);
- brw_inst_set_exec_size(devinfo, insn,
- brw_inst_exec_size(devinfo, p->current));
+ brw_inst_set_exec_size(devinfo, insn, brw_get_default_exec_size(p));
return insn;
}
@@ -1671,8 +1668,7 @@ gen6_HALT(struct brw_codegen *p)
}
brw_inst_set_qtr_control(devinfo, insn, BRW_COMPRESSION_NONE);
- brw_inst_set_exec_size(devinfo, insn,
- brw_inst_exec_size(devinfo, p->current));
+ brw_inst_set_exec_size(devinfo, insn, brw_get_default_exec_size(p));
return insn;
}
@@ -1778,8 +1774,7 @@ brw_WHILE(struct brw_codegen *p)
brw_set_src1(p, insn, retype(brw_null_reg(), BRW_REGISTER_TYPE_D));
}
- brw_inst_set_exec_size(devinfo, insn,
- brw_inst_exec_size(devinfo, p->current));
+ brw_inst_set_exec_size(devinfo, insn, brw_get_default_exec_size(p));
} else {
if (p->single_program_flow) {
@@ -2207,7 +2202,7 @@ void brw_oword_block_read(struct brw_codegen *p,
const unsigned target_cache =
(devinfo->gen >= 6 ? GEN6_SFID_DATAPORT_CONSTANT_CACHE :
BRW_DATAPORT_READ_TARGET_DATA_CACHE);
- const unsigned exec_size = 1 << brw_inst_exec_size(devinfo, p->current);
+ const unsigned exec_size = 1 << brw_get_default_exec_size(p);
/* On newer hardware, offset is in units of owords. */
if (devinfo->gen >= 6)
@@ -2277,7 +2272,7 @@ void brw_fb_WRITE(struct brw_codegen *p,
unsigned msg_type;
struct brw_reg dest, src0;
- if (brw_inst_exec_size(devinfo, p->current) >= BRW_EXECUTE_16)
+ if (brw_get_default_exec_size(p) >= BRW_EXECUTE_16)
dest = retype(vec16(brw_null_reg()), BRW_REGISTER_TYPE_UW);
else
dest = retype(vec8(brw_null_reg()), BRW_REGISTER_TYPE_UW);
@@ -2330,7 +2325,7 @@ gen9_fb_READ(struct brw_codegen *p,
const struct gen_device_info *devinfo = p->devinfo;
assert(devinfo->gen >= 9);
const unsigned msg_subtype =
- brw_inst_exec_size(devinfo, p->current) == BRW_EXECUTE_16 ? 0 : 1;
+ brw_get_default_exec_size(p) == BRW_EXECUTE_16 ? 0 : 1;
brw_inst *insn = next_insn(p, BRW_OPCODE_SENDC);
brw_set_dest(p, insn, dst);
@@ -2341,8 +2336,7 @@ gen9_fb_READ(struct brw_codegen *p,
GEN6_SFID_DATAPORT_RENDER_CACHE,
msg_length, true /* header_present */,
response_length);
- brw_inst_set_rt_slot_group(devinfo, insn,
- brw_inst_qtr_control(devinfo, p->current) / 2);
+ brw_inst_set_rt_slot_group(devinfo, insn, brw_get_default_group(p) / 16);
return insn;
}
@@ -2837,11 +2831,9 @@ brw_surface_payload_size(struct brw_codegen *p,
bool has_simd4x2,
bool has_simd16)
{
- if (has_simd4x2 &&
- brw_inst_access_mode(p->devinfo, p->current) == BRW_ALIGN_16)
+ if (has_simd4x2 && brw_get_default_access_mode(p) == BRW_ALIGN_16)
return 1;
- else if (has_simd16 &&
- brw_inst_exec_size(p->devinfo, p->current) == BRW_EXECUTE_16)
+ else if (has_simd16 && brw_get_default_exec_size(p) == BRW_EXECUTE_16)
return 2 * num_channels;
else
return num_channels;
@@ -2859,8 +2851,8 @@ brw_set_dp_untyped_atomic_message(struct brw_codegen *p,
(response_expected ? 1 << 5 : 0); /* Return data expected */
if (devinfo->gen >= 8 || devinfo->is_haswell) {
- if (brw_inst_access_mode(devinfo, p->current) == BRW_ALIGN_1) {
- if (brw_inst_exec_size(devinfo, p->current) != BRW_EXECUTE_16)
+ if (brw_get_default_access_mode(p) == BRW_ALIGN_1) {
+ if (brw_get_default_exec_size(p) != BRW_EXECUTE_16)
msg_control |= 1 << 4; /* SIMD8 mode */
brw_inst_set_dp_msg_type(devinfo, insn,
@@ -2873,7 +2865,7 @@ brw_set_dp_untyped_atomic_message(struct brw_codegen *p,
brw_inst_set_dp_msg_type(devinfo, insn,
GEN7_DATAPORT_DC_UNTYPED_ATOMIC_OP);
- if (brw_inst_exec_size(devinfo, p->current) != BRW_EXECUTE_16)
+ if (brw_get_default_exec_size(p) != BRW_EXECUTE_16)
msg_control |= 1 << 4; /* SIMD8 mode */
}
@@ -2894,7 +2886,7 @@ brw_untyped_atomic(struct brw_codegen *p,
const unsigned sfid = (devinfo->gen >= 8 || devinfo->is_haswell ?
HSW_SFID_DATAPORT_DATA_CACHE_1 :
GEN7_SFID_DATAPORT_DATA_CACHE);
- const bool align1 = brw_inst_access_mode(devinfo, p->current) == BRW_ALIGN_1;
+ const bool align1 = brw_get_default_access_mode(p) == BRW_ALIGN_1;
/* Mask out unused components -- This is especially important in Align16
* mode on generations that don't have native support for SIMD4x2 atomics,
* because unused but enabled components will cause the dataport to perform
@@ -2921,8 +2913,8 @@ brw_set_dp_untyped_surface_read_message(struct brw_codegen *p,
/* Set mask of 32-bit channels to drop. */
unsigned msg_control = 0xf & (0xf << num_channels);
- if (brw_inst_access_mode(devinfo, p->current) == BRW_ALIGN_1) {
- if (brw_inst_exec_size(devinfo, p->current) == BRW_EXECUTE_16)
+ if (brw_get_default_access_mode(p) == BRW_ALIGN_1) {
+ if (brw_get_default_exec_size(p) == BRW_EXECUTE_16)
msg_control |= 1 << 4; /* SIMD16 mode */
else
msg_control |= 2 << 4; /* SIMD8 mode */
@@ -2965,8 +2957,8 @@ brw_set_dp_untyped_surface_write_message(struct brw_codegen *p,
/* Set mask of 32-bit channels to drop. */
unsigned msg_control = 0xf & (0xf << num_channels);
- if (brw_inst_access_mode(devinfo, p->current) == BRW_ALIGN_1) {
- if (brw_inst_exec_size(devinfo, p->current) == BRW_EXECUTE_16)
+ if (brw_get_default_access_mode(p) == BRW_ALIGN_1) {
+ if (brw_get_default_exec_size(p) == BRW_EXECUTE_16)
msg_control |= 1 << 4; /* SIMD16 mode */
else
msg_control |= 2 << 4; /* SIMD8 mode */
@@ -2996,7 +2988,7 @@ brw_untyped_surface_write(struct brw_codegen *p,
const unsigned sfid = (devinfo->gen >= 8 || devinfo->is_haswell ?
HSW_SFID_DATAPORT_DATA_CACHE_1 :
GEN7_SFID_DATAPORT_DATA_CACHE);
- const bool align1 = brw_inst_access_mode(devinfo, p->current) == BRW_ALIGN_1;
+ const bool align1 = brw_get_default_access_mode(p) == BRW_ALIGN_1;
/* Mask out unused components -- See comment in brw_untyped_atomic(). */
const unsigned mask = devinfo->gen == 7 && !devinfo->is_haswell && !align1 ?
WRITEMASK_X : WRITEMASK_XYZW;
@@ -3034,7 +3026,7 @@ brw_byte_scattered_read(struct brw_codegen *p,
{
const struct gen_device_info *devinfo = p->devinfo;
assert(devinfo->gen > 7 || devinfo->is_haswell);
- assert(brw_inst_access_mode(devinfo, p->current) == BRW_ALIGN_1);
+ assert(brw_get_default_access_mode(p) == BRW_ALIGN_1);
const unsigned sfid = GEN7_SFID_DATAPORT_DATA_CACHE;
struct brw_inst *insn = brw_send_indirect_surface_message(
@@ -3045,7 +3037,7 @@ brw_byte_scattered_read(struct brw_codegen *p,
unsigned msg_control =
brw_byte_scattered_data_element_from_bit_size(bit_size) << 2;
- if (brw_inst_exec_size(devinfo, p->current) == BRW_EXECUTE_16)
+ if (brw_get_default_exec_size(p) == BRW_EXECUTE_16)
msg_control |= 1; /* SIMD16 mode */
else
msg_control |= 0; /* SIMD8 mode */
@@ -3065,7 +3057,7 @@ brw_byte_scattered_write(struct brw_codegen *p,
{
const struct gen_device_info *devinfo = p->devinfo;
assert(devinfo->gen > 7 || devinfo->is_haswell);
- assert(brw_inst_access_mode(devinfo, p->current) == BRW_ALIGN_1);
+ assert(brw_get_default_access_mode(p) == BRW_ALIGN_1);
const unsigned sfid = GEN7_SFID_DATAPORT_DATA_CACHE;
struct brw_inst *insn = brw_send_indirect_surface_message(
@@ -3075,7 +3067,7 @@ brw_byte_scattered_write(struct brw_codegen *p,
unsigned msg_control =
brw_byte_scattered_data_element_from_bit_size(bit_size) << 2;
- if (brw_inst_exec_size(devinfo, p->current) == BRW_EXECUTE_16)
+ if (brw_get_default_exec_size(p) == BRW_EXECUTE_16)
msg_control |= 1;
else
msg_control |= 0;
@@ -3097,8 +3089,8 @@ brw_set_dp_typed_atomic_message(struct brw_codegen *p,
(response_expected ? 1 << 5 : 0); /* Return data expected */
if (devinfo->gen >= 8 || devinfo->is_haswell) {
- if (brw_inst_access_mode(devinfo, p->current) == BRW_ALIGN_1) {
- if (brw_inst_qtr_control(devinfo, p->current) % 2 == 1)
+ if (brw_get_default_access_mode(p) == BRW_ALIGN_1) {
+ if ((brw_get_default_group(p) / 8) % 2 == 1)
msg_control |= 1 << 4; /* Use high 8 slots of the sample mask */
brw_inst_set_dp_msg_type(devinfo, insn,
@@ -3112,7 +3104,7 @@ brw_set_dp_typed_atomic_message(struct brw_codegen *p,
brw_inst_set_dp_msg_type(devinfo, insn,
GEN7_DATAPORT_RC_TYPED_ATOMIC_OP);
- if (brw_inst_qtr_control(devinfo, p->current) % 2 == 1)
+ if ((brw_get_default_group(p) / 8) % 2 == 1)
msg_control |= 1 << 4; /* Use high 8 slots of the sample mask */
}
@@ -3132,7 +3124,7 @@ brw_typed_atomic(struct brw_codegen *p,
const unsigned sfid = (devinfo->gen >= 8 || devinfo->is_haswell ?
HSW_SFID_DATAPORT_DATA_CACHE_1 :
GEN6_SFID_DATAPORT_RENDER_CACHE);
- const bool align1 = (brw_inst_access_mode(devinfo, p->current) == BRW_ALIGN_1);
+ const bool align1 = brw_get_default_access_mode(p) == BRW_ALIGN_1;
/* Mask out unused components -- See comment in brw_untyped_atomic(). */
const unsigned mask = align1 ? WRITEMASK_XYZW : WRITEMASK_X;
struct brw_inst *insn = brw_send_indirect_surface_message(
@@ -3155,8 +3147,8 @@ brw_set_dp_typed_surface_read_message(struct brw_codegen *p,
unsigned msg_control = 0xf & (0xf << num_channels);
if (devinfo->gen >= 8 || devinfo->is_haswell) {
- if (brw_inst_access_mode(devinfo, p->current) == BRW_ALIGN_1) {
- if (brw_inst_qtr_control(devinfo, p->current) % 2 == 1)
+ if (brw_get_default_access_mode(p) == BRW_ALIGN_1) {
+ if ((brw_get_default_group(p) / 8) % 2 == 1)
msg_control |= 2 << 4; /* Use high 8 slots of the sample mask */
else
msg_control |= 1 << 4; /* Use low 8 slots of the sample mask */
@@ -3165,8 +3157,8 @@ brw_set_dp_typed_surface_read_message(struct brw_codegen *p,
brw_inst_set_dp_msg_type(devinfo, insn,
HSW_DATAPORT_DC_PORT1_TYPED_SURFACE_READ);
} else {
- if (brw_inst_access_mode(devinfo, p->current) == BRW_ALIGN_1) {
- if (brw_inst_qtr_control(devinfo, p->current) % 2 == 1)
+ if (brw_get_default_access_mode(p) == BRW_ALIGN_1) {
+ if ((brw_get_default_group(p) / 8) % 2 == 1)
msg_control |= 1 << 5; /* Use high 8 slots of the sample mask */
}
@@ -3210,8 +3202,8 @@ brw_set_dp_typed_surface_write_message(struct brw_codegen *p,
unsigned msg_control = 0xf & (0xf << num_channels);
if (devinfo->gen >= 8 || devinfo->is_haswell) {
- if (brw_inst_access_mode(devinfo, p->current) == BRW_ALIGN_1) {
- if (brw_inst_qtr_control(devinfo, p->current) % 2 == 1)
+ if (brw_get_default_access_mode(p) == BRW_ALIGN_1) {
+ if ((brw_get_default_group(p) / 8) % 2 == 1)
msg_control |= 2 << 4; /* Use high 8 slots of the sample mask */
else
msg_control |= 1 << 4; /* Use low 8 slots of the sample mask */
@@ -3221,8 +3213,8 @@ brw_set_dp_typed_surface_write_message(struct brw_codegen *p,
HSW_DATAPORT_DC_PORT1_TYPED_SURFACE_WRITE);
} else {
- if (brw_inst_access_mode(devinfo, p->current) == BRW_ALIGN_1) {
- if (brw_inst_qtr_control(devinfo, p->current) % 2 == 1)
+ if (brw_get_default_access_mode(p) == BRW_ALIGN_1) {
+ if ((brw_get_default_group(p) / 8) % 2 == 1)
msg_control |= 1 << 5; /* Use high 8 slots of the sample mask */
}
@@ -3245,7 +3237,7 @@ brw_typed_surface_write(struct brw_codegen *p,
const unsigned sfid = (devinfo->gen >= 8 || devinfo->is_haswell ?
HSW_SFID_DATAPORT_DATA_CACHE_1 :
GEN6_SFID_DATAPORT_RENDER_CACHE);
- const bool align1 = (brw_inst_access_mode(devinfo, p->current) == BRW_ALIGN_1);
+ const bool align1 = brw_get_default_access_mode(p) == BRW_ALIGN_1;
/* Mask out unused components -- See comment in brw_untyped_atomic(). */
const unsigned mask = (devinfo->gen == 7 && !devinfo->is_haswell && !align1 ?
WRITEMASK_X : WRITEMASK_XYZW);
@@ -3346,7 +3338,7 @@ brw_pixel_interpolator_query(struct brw_codegen *p,
{
const struct gen_device_info *devinfo = p->devinfo;
struct brw_inst *insn;
- const uint16_t exec_size = brw_inst_exec_size(devinfo, p->current);
+ const uint16_t exec_size = brw_get_default_exec_size(p);
/* brw_send_indirect_message will automatically use a direct send message
* if data is actually immediate.
@@ -3370,8 +3362,8 @@ brw_find_live_channel(struct brw_codegen *p, struct brw_reg dst,
struct brw_reg mask)
{
const struct gen_device_info *devinfo = p->devinfo;
- const unsigned exec_size = 1 << brw_inst_exec_size(devinfo, p->current);
- const unsigned qtr_control = brw_inst_qtr_control(devinfo, p->current);
+ const unsigned exec_size = 1 << brw_get_default_exec_size(p);
+ const unsigned qtr_control = brw_get_default_group(p) / 8;
brw_inst *inst;
assert(devinfo->gen >= 7);
@@ -3379,7 +3371,7 @@ brw_find_live_channel(struct brw_codegen *p, struct brw_reg dst,
brw_push_insn_state(p);
- if (brw_inst_access_mode(devinfo, p->current) == BRW_ALIGN_1) {
+ if (brw_get_default_access_mode(p) == BRW_ALIGN_1) {
brw_set_default_mask_control(p, BRW_MASK_DISABLE);
if (devinfo->gen >= 8) {
@@ -3486,7 +3478,7 @@ brw_broadcast(struct brw_codegen *p,
struct brw_reg idx)
{
const struct gen_device_info *devinfo = p->devinfo;
- const bool align1 = brw_inst_access_mode(devinfo, p->current) == BRW_ALIGN_1;
+ const bool align1 = brw_get_default_access_mode(p) == BRW_ALIGN_1;
brw_inst *inst;
brw_push_insn_state(p);
diff --git a/src/intel/compiler/brw_fs_generator.cpp b/src/intel/compiler/brw_fs_generator.cpp
index de84f468d9b..09839d0b4da 100644
--- a/src/intel/compiler/brw_fs_generator.cpp
+++ b/src/intel/compiler/brw_fs_generator.cpp
@@ -2042,7 +2042,7 @@ fs_generator::generate_code(const cfg_t *cfg, int dispatch_width)
assert(devinfo->gen == 6);
gen6_IF(p, inst->conditional_mod, src[0], src[1]);
} else {
- brw_IF(p, brw_inst_exec_size(devinfo, p->current));
+ brw_IF(p, brw_get_default_exec_size(p));
}
break;
@@ -2054,7 +2054,7 @@ fs_generator::generate_code(const cfg_t *cfg, int dispatch_width)
break;
case BRW_OPCODE_DO:
- brw_DO(p, brw_inst_exec_size(devinfo, p->current));
+ brw_DO(p, brw_get_default_exec_size(p));
break;
case BRW_OPCODE_BREAK:
generated by cgit v1.2.3 (git 2.39.1) at 2024-12-18 12:29:07 +0000