mesa/src/compiler/nir/nir_validate.c

1238 lines
39 KiB
C
Raw Normal View History

/*
* Copyright © 2014 Intel Corporation
*
* Permission is hereby granted, free of charge, to any person obtaining a
* copy of this software and associated documentation files (the "Software"),
* to deal in the Software without restriction, including without limitation
* the rights to use, copy, modify, merge, publish, distribute, sublicense,
* and/or sell copies of the Software, and to permit persons to whom the
* Software is furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice (including the next
* paragraph) shall be included in all copies or substantial portions of the
* Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
* THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
* FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
* IN THE SOFTWARE.
*
* Authors:
* Connor Abbott (cwabbott0@gmail.com)
*
*/
#include "nir.h"
#include <assert.h>
/*
* This file checks for invalid IR indicating a bug somewhere in the compiler.
*/
/* Since this file is just a pile of asserts, don't bother compiling it if
* we're not building a debug build.
*/
#ifndef NDEBUG
/*
* Per-register validation state.
*/
typedef struct {
/*
* equivalent to the uses and defs in nir_register, but built up by the
* validator. At the end, we verify that the sets have the same entries.
*/
struct set *uses, *if_uses, *defs;
nir_function_impl *where_defined; /* NULL for global registers */
} reg_validate_state;
typedef struct {
/*
* equivalent to the uses in nir_ssa_def, but built up by the validator.
* At the end, we verify that the sets have the same entries.
*/
struct set *uses, *if_uses;
nir_function_impl *where_defined;
} ssa_def_validate_state;
typedef struct {
/* map of register -> validation state (struct above) */
struct hash_table *regs;
/* the current shader being validated */
nir_shader *shader;
/* the current instruction being validated */
nir_instr *instr;
/* the current variable being validated */
nir_variable *var;
/* the current basic block being validated */
nir_block *block;
/* the current if statement being validated */
nir_if *if_stmt;
/* the current loop being visited */
nir_loop *loop;
/* the parent of the current cf node being visited */
nir_cf_node *parent_node;
/* the current function implementation being validated */
nir_function_impl *impl;
/* map of SSA value -> function implementation where it is defined */
struct hash_table *ssa_defs;
/* bitset of ssa definitions we have found; used to check uniqueness */
BITSET_WORD *ssa_defs_found;
/* bitset of registers we have currently found; used to check uniqueness */
BITSET_WORD *regs_found;
/* map of variable -> function implementation where it is defined or NULL
* if it is a global variable
*/
struct hash_table *var_defs;
/* map of instruction/var/etc to failed assert string */
struct hash_table *errors;
} validate_state;
static void
log_error(validate_state *state, const char *cond, const char *file, int line)
{
const void *obj;
if (state->instr)
obj = state->instr;
else if (state->var)
obj = state->var;
else
obj = cond;
char *msg = ralloc_asprintf(state->errors, "error: %s (%s:%d)",
cond, file, line);
_mesa_hash_table_insert(state->errors, obj, msg);
}
#define validate_assert(state, cond) do { \
if (!(cond)) \
log_error(state, #cond, __FILE__, __LINE__); \
} while (0)
static void validate_src(nir_src *src, validate_state *state,
unsigned bit_size, unsigned num_components);
static void
validate_reg_src(nir_src *src, validate_state *state,
unsigned bit_size, unsigned num_components)
{
validate_assert(state, src->reg.reg != NULL);
struct hash_entry *entry;
nir/nir: Use a linked list instead of a hash set for use/def sets This commit switches us from the current setup of using hash sets for use/def sets to using linked lists. Doing so should save us quite a bit of memory because we aren't carrying around 3 hash sets per register and 2 per SSA value. It should also save us CPU time because adding/removing things from use/def sets is 4 pointer manipulations instead of a hash lookup. Running shader-db 50 times with USE_NIR=0, NIR, and NIR + use/def lists: GLSL IR Only: 586.4 +/- 1.653833 NIR with hash sets: 675.4 +/- 2.502108 NIR + use/def lists: 641.2 +/- 1.557043 I also ran a memory usage experiment with Ken's patch to delete GLSL IR and keep NIR. This patch cuts an aditional 42.9 MiB of ralloc'd memory over and above what we gained by deleting the GLSL IR on the same dota trace. On the code complexity side of things, some things are now much easier and others are a bit harder. One of the operations we perform constantly in optimization passes is to replace one source with another. Due to the fact that an instruction can use the same SSA value multiple times, we had to iterate through the sources of the instruction and determine if the use we were replacing was the only one before removing it from the set of uses. With this patch, uses are per-source not per-instruction so we can just remove it safely. On the other hand, trying to iterate over all of the instructions that use a given value is more difficult. Fortunately, the two places we do that are the ffma peephole where it doesn't matter and GCM where we already gracefully handle duplicates visits to an instruction. Another aspect here is that using linked lists in this way can be tricky to get right. With sets, things were quite forgiving and the worst that happened if you didn't properly remove a use was that it would get caught in the validator. With linked lists, it can lead to linked list corruption which can be harder to track. However, we do just as much validation of the linked lists as we did of the sets so the validator should still catch these problems. While working on this series, the vast majority of the bugs I had to fix were caught by assertions. I don't think the lists are going to be that much worse than the sets. Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
2015-04-24 10:16:27 -07:00
entry = _mesa_hash_table_search(state->regs, src->reg.reg);
validate_assert(state, entry);
reg_validate_state *reg_state = (reg_validate_state *) entry->data;
if (state->instr) {
nir/nir: Use a linked list instead of a hash set for use/def sets This commit switches us from the current setup of using hash sets for use/def sets to using linked lists. Doing so should save us quite a bit of memory because we aren't carrying around 3 hash sets per register and 2 per SSA value. It should also save us CPU time because adding/removing things from use/def sets is 4 pointer manipulations instead of a hash lookup. Running shader-db 50 times with USE_NIR=0, NIR, and NIR + use/def lists: GLSL IR Only: 586.4 +/- 1.653833 NIR with hash sets: 675.4 +/- 2.502108 NIR + use/def lists: 641.2 +/- 1.557043 I also ran a memory usage experiment with Ken's patch to delete GLSL IR and keep NIR. This patch cuts an aditional 42.9 MiB of ralloc'd memory over and above what we gained by deleting the GLSL IR on the same dota trace. On the code complexity side of things, some things are now much easier and others are a bit harder. One of the operations we perform constantly in optimization passes is to replace one source with another. Due to the fact that an instruction can use the same SSA value multiple times, we had to iterate through the sources of the instruction and determine if the use we were replacing was the only one before removing it from the set of uses. With this patch, uses are per-source not per-instruction so we can just remove it safely. On the other hand, trying to iterate over all of the instructions that use a given value is more difficult. Fortunately, the two places we do that are the ffma peephole where it doesn't matter and GCM where we already gracefully handle duplicates visits to an instruction. Another aspect here is that using linked lists in this way can be tricky to get right. With sets, things were quite forgiving and the worst that happened if you didn't properly remove a use was that it would get caught in the validator. With linked lists, it can lead to linked list corruption which can be harder to track. However, we do just as much validation of the linked lists as we did of the sets so the validator should still catch these problems. While working on this series, the vast majority of the bugs I had to fix were caught by assertions. I don't think the lists are going to be that much worse than the sets. Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
2015-04-24 10:16:27 -07:00
_mesa_set_add(reg_state->uses, src);
} else {
validate_assert(state, state->if_stmt);
nir/nir: Use a linked list instead of a hash set for use/def sets This commit switches us from the current setup of using hash sets for use/def sets to using linked lists. Doing so should save us quite a bit of memory because we aren't carrying around 3 hash sets per register and 2 per SSA value. It should also save us CPU time because adding/removing things from use/def sets is 4 pointer manipulations instead of a hash lookup. Running shader-db 50 times with USE_NIR=0, NIR, and NIR + use/def lists: GLSL IR Only: 586.4 +/- 1.653833 NIR with hash sets: 675.4 +/- 2.502108 NIR + use/def lists: 641.2 +/- 1.557043 I also ran a memory usage experiment with Ken's patch to delete GLSL IR and keep NIR. This patch cuts an aditional 42.9 MiB of ralloc'd memory over and above what we gained by deleting the GLSL IR on the same dota trace. On the code complexity side of things, some things are now much easier and others are a bit harder. One of the operations we perform constantly in optimization passes is to replace one source with another. Due to the fact that an instruction can use the same SSA value multiple times, we had to iterate through the sources of the instruction and determine if the use we were replacing was the only one before removing it from the set of uses. With this patch, uses are per-source not per-instruction so we can just remove it safely. On the other hand, trying to iterate over all of the instructions that use a given value is more difficult. Fortunately, the two places we do that are the ffma peephole where it doesn't matter and GCM where we already gracefully handle duplicates visits to an instruction. Another aspect here is that using linked lists in this way can be tricky to get right. With sets, things were quite forgiving and the worst that happened if you didn't properly remove a use was that it would get caught in the validator. With linked lists, it can lead to linked list corruption which can be harder to track. However, we do just as much validation of the linked lists as we did of the sets so the validator should still catch these problems. While working on this series, the vast majority of the bugs I had to fix were caught by assertions. I don't think the lists are going to be that much worse than the sets. Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
2015-04-24 10:16:27 -07:00
_mesa_set_add(reg_state->if_uses, src);
}
nir/nir: Use a linked list instead of a hash set for use/def sets This commit switches us from the current setup of using hash sets for use/def sets to using linked lists. Doing so should save us quite a bit of memory because we aren't carrying around 3 hash sets per register and 2 per SSA value. It should also save us CPU time because adding/removing things from use/def sets is 4 pointer manipulations instead of a hash lookup. Running shader-db 50 times with USE_NIR=0, NIR, and NIR + use/def lists: GLSL IR Only: 586.4 +/- 1.653833 NIR with hash sets: 675.4 +/- 2.502108 NIR + use/def lists: 641.2 +/- 1.557043 I also ran a memory usage experiment with Ken's patch to delete GLSL IR and keep NIR. This patch cuts an aditional 42.9 MiB of ralloc'd memory over and above what we gained by deleting the GLSL IR on the same dota trace. On the code complexity side of things, some things are now much easier and others are a bit harder. One of the operations we perform constantly in optimization passes is to replace one source with another. Due to the fact that an instruction can use the same SSA value multiple times, we had to iterate through the sources of the instruction and determine if the use we were replacing was the only one before removing it from the set of uses. With this patch, uses are per-source not per-instruction so we can just remove it safely. On the other hand, trying to iterate over all of the instructions that use a given value is more difficult. Fortunately, the two places we do that are the ffma peephole where it doesn't matter and GCM where we already gracefully handle duplicates visits to an instruction. Another aspect here is that using linked lists in this way can be tricky to get right. With sets, things were quite forgiving and the worst that happened if you didn't properly remove a use was that it would get caught in the validator. With linked lists, it can lead to linked list corruption which can be harder to track. However, we do just as much validation of the linked lists as we did of the sets so the validator should still catch these problems. While working on this series, the vast majority of the bugs I had to fix were caught by assertions. I don't think the lists are going to be that much worse than the sets. Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
2015-04-24 10:16:27 -07:00
if (!src->reg.reg->is_global) {
validate_assert(state, reg_state->where_defined == state->impl &&
"using a register declared in a different function");
}
if (!src->reg.reg->is_packed) {
if (bit_size)
validate_assert(state, src->reg.reg->bit_size == bit_size);
if (num_components)
validate_assert(state, src->reg.reg->num_components == num_components);
}
validate_assert(state, (src->reg.reg->num_array_elems == 0 ||
nir/nir: Use a linked list instead of a hash set for use/def sets This commit switches us from the current setup of using hash sets for use/def sets to using linked lists. Doing so should save us quite a bit of memory because we aren't carrying around 3 hash sets per register and 2 per SSA value. It should also save us CPU time because adding/removing things from use/def sets is 4 pointer manipulations instead of a hash lookup. Running shader-db 50 times with USE_NIR=0, NIR, and NIR + use/def lists: GLSL IR Only: 586.4 +/- 1.653833 NIR with hash sets: 675.4 +/- 2.502108 NIR + use/def lists: 641.2 +/- 1.557043 I also ran a memory usage experiment with Ken's patch to delete GLSL IR and keep NIR. This patch cuts an aditional 42.9 MiB of ralloc'd memory over and above what we gained by deleting the GLSL IR on the same dota trace. On the code complexity side of things, some things are now much easier and others are a bit harder. One of the operations we perform constantly in optimization passes is to replace one source with another. Due to the fact that an instruction can use the same SSA value multiple times, we had to iterate through the sources of the instruction and determine if the use we were replacing was the only one before removing it from the set of uses. With this patch, uses are per-source not per-instruction so we can just remove it safely. On the other hand, trying to iterate over all of the instructions that use a given value is more difficult. Fortunately, the two places we do that are the ffma peephole where it doesn't matter and GCM where we already gracefully handle duplicates visits to an instruction. Another aspect here is that using linked lists in this way can be tricky to get right. With sets, things were quite forgiving and the worst that happened if you didn't properly remove a use was that it would get caught in the validator. With linked lists, it can lead to linked list corruption which can be harder to track. However, we do just as much validation of the linked lists as we did of the sets so the validator should still catch these problems. While working on this series, the vast majority of the bugs I had to fix were caught by assertions. I don't think the lists are going to be that much worse than the sets. Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
2015-04-24 10:16:27 -07:00
src->reg.base_offset < src->reg.reg->num_array_elems) &&
"definitely out-of-bounds array access");
nir/nir: Use a linked list instead of a hash set for use/def sets This commit switches us from the current setup of using hash sets for use/def sets to using linked lists. Doing so should save us quite a bit of memory because we aren't carrying around 3 hash sets per register and 2 per SSA value. It should also save us CPU time because adding/removing things from use/def sets is 4 pointer manipulations instead of a hash lookup. Running shader-db 50 times with USE_NIR=0, NIR, and NIR + use/def lists: GLSL IR Only: 586.4 +/- 1.653833 NIR with hash sets: 675.4 +/- 2.502108 NIR + use/def lists: 641.2 +/- 1.557043 I also ran a memory usage experiment with Ken's patch to delete GLSL IR and keep NIR. This patch cuts an aditional 42.9 MiB of ralloc'd memory over and above what we gained by deleting the GLSL IR on the same dota trace. On the code complexity side of things, some things are now much easier and others are a bit harder. One of the operations we perform constantly in optimization passes is to replace one source with another. Due to the fact that an instruction can use the same SSA value multiple times, we had to iterate through the sources of the instruction and determine if the use we were replacing was the only one before removing it from the set of uses. With this patch, uses are per-source not per-instruction so we can just remove it safely. On the other hand, trying to iterate over all of the instructions that use a given value is more difficult. Fortunately, the two places we do that are the ffma peephole where it doesn't matter and GCM where we already gracefully handle duplicates visits to an instruction. Another aspect here is that using linked lists in this way can be tricky to get right. With sets, things were quite forgiving and the worst that happened if you didn't properly remove a use was that it would get caught in the validator. With linked lists, it can lead to linked list corruption which can be harder to track. However, we do just as much validation of the linked lists as we did of the sets so the validator should still catch these problems. While working on this series, the vast majority of the bugs I had to fix were caught by assertions. I don't think the lists are going to be that much worse than the sets. Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
2015-04-24 10:16:27 -07:00
if (src->reg.indirect) {
validate_assert(state, src->reg.reg->num_array_elems != 0);
validate_assert(state, (src->reg.indirect->is_ssa ||
nir/nir: Use a linked list instead of a hash set for use/def sets This commit switches us from the current setup of using hash sets for use/def sets to using linked lists. Doing so should save us quite a bit of memory because we aren't carrying around 3 hash sets per register and 2 per SSA value. It should also save us CPU time because adding/removing things from use/def sets is 4 pointer manipulations instead of a hash lookup. Running shader-db 50 times with USE_NIR=0, NIR, and NIR + use/def lists: GLSL IR Only: 586.4 +/- 1.653833 NIR with hash sets: 675.4 +/- 2.502108 NIR + use/def lists: 641.2 +/- 1.557043 I also ran a memory usage experiment with Ken's patch to delete GLSL IR and keep NIR. This patch cuts an aditional 42.9 MiB of ralloc'd memory over and above what we gained by deleting the GLSL IR on the same dota trace. On the code complexity side of things, some things are now much easier and others are a bit harder. One of the operations we perform constantly in optimization passes is to replace one source with another. Due to the fact that an instruction can use the same SSA value multiple times, we had to iterate through the sources of the instruction and determine if the use we were replacing was the only one before removing it from the set of uses. With this patch, uses are per-source not per-instruction so we can just remove it safely. On the other hand, trying to iterate over all of the instructions that use a given value is more difficult. Fortunately, the two places we do that are the ffma peephole where it doesn't matter and GCM where we already gracefully handle duplicates visits to an instruction. Another aspect here is that using linked lists in this way can be tricky to get right. With sets, things were quite forgiving and the worst that happened if you didn't properly remove a use was that it would get caught in the validator. With linked lists, it can lead to linked list corruption which can be harder to track. However, we do just as much validation of the linked lists as we did of the sets so the validator should still catch these problems. While working on this series, the vast majority of the bugs I had to fix were caught by assertions. I don't think the lists are going to be that much worse than the sets. Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
2015-04-24 10:16:27 -07:00
src->reg.indirect->reg.indirect == NULL) &&
"only one level of indirection allowed");
validate_src(src->reg.indirect, state, 32, 1);
}
}
static void
validate_ssa_src(nir_src *src, validate_state *state,
unsigned bit_size, unsigned num_components)
{
validate_assert(state, src->ssa != NULL);
nir/nir: Use a linked list instead of a hash set for use/def sets This commit switches us from the current setup of using hash sets for use/def sets to using linked lists. Doing so should save us quite a bit of memory because we aren't carrying around 3 hash sets per register and 2 per SSA value. It should also save us CPU time because adding/removing things from use/def sets is 4 pointer manipulations instead of a hash lookup. Running shader-db 50 times with USE_NIR=0, NIR, and NIR + use/def lists: GLSL IR Only: 586.4 +/- 1.653833 NIR with hash sets: 675.4 +/- 2.502108 NIR + use/def lists: 641.2 +/- 1.557043 I also ran a memory usage experiment with Ken's patch to delete GLSL IR and keep NIR. This patch cuts an aditional 42.9 MiB of ralloc'd memory over and above what we gained by deleting the GLSL IR on the same dota trace. On the code complexity side of things, some things are now much easier and others are a bit harder. One of the operations we perform constantly in optimization passes is to replace one source with another. Due to the fact that an instruction can use the same SSA value multiple times, we had to iterate through the sources of the instruction and determine if the use we were replacing was the only one before removing it from the set of uses. With this patch, uses are per-source not per-instruction so we can just remove it safely. On the other hand, trying to iterate over all of the instructions that use a given value is more difficult. Fortunately, the two places we do that are the ffma peephole where it doesn't matter and GCM where we already gracefully handle duplicates visits to an instruction. Another aspect here is that using linked lists in this way can be tricky to get right. With sets, things were quite forgiving and the worst that happened if you didn't properly remove a use was that it would get caught in the validator. With linked lists, it can lead to linked list corruption which can be harder to track. However, we do just as much validation of the linked lists as we did of the sets so the validator should still catch these problems. While working on this series, the vast majority of the bugs I had to fix were caught by assertions. I don't think the lists are going to be that much worse than the sets. Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
2015-04-24 10:16:27 -07:00
struct hash_entry *entry = _mesa_hash_table_search(state->ssa_defs, src->ssa);
validate_assert(state, entry);
if (!entry)
return;
ssa_def_validate_state *def_state = (ssa_def_validate_state *)entry->data;
validate_assert(state, def_state->where_defined == state->impl &&
"using an SSA value defined in a different function");
if (state->instr) {
nir/nir: Use a linked list instead of a hash set for use/def sets This commit switches us from the current setup of using hash sets for use/def sets to using linked lists. Doing so should save us quite a bit of memory because we aren't carrying around 3 hash sets per register and 2 per SSA value. It should also save us CPU time because adding/removing things from use/def sets is 4 pointer manipulations instead of a hash lookup. Running shader-db 50 times with USE_NIR=0, NIR, and NIR + use/def lists: GLSL IR Only: 586.4 +/- 1.653833 NIR with hash sets: 675.4 +/- 2.502108 NIR + use/def lists: 641.2 +/- 1.557043 I also ran a memory usage experiment with Ken's patch to delete GLSL IR and keep NIR. This patch cuts an aditional 42.9 MiB of ralloc'd memory over and above what we gained by deleting the GLSL IR on the same dota trace. On the code complexity side of things, some things are now much easier and others are a bit harder. One of the operations we perform constantly in optimization passes is to replace one source with another. Due to the fact that an instruction can use the same SSA value multiple times, we had to iterate through the sources of the instruction and determine if the use we were replacing was the only one before removing it from the set of uses. With this patch, uses are per-source not per-instruction so we can just remove it safely. On the other hand, trying to iterate over all of the instructions that use a given value is more difficult. Fortunately, the two places we do that are the ffma peephole where it doesn't matter and GCM where we already gracefully handle duplicates visits to an instruction. Another aspect here is that using linked lists in this way can be tricky to get right. With sets, things were quite forgiving and the worst that happened if you didn't properly remove a use was that it would get caught in the validator. With linked lists, it can lead to linked list corruption which can be harder to track. However, we do just as much validation of the linked lists as we did of the sets so the validator should still catch these problems. While working on this series, the vast majority of the bugs I had to fix were caught by assertions. I don't think the lists are going to be that much worse than the sets. Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
2015-04-24 10:16:27 -07:00
_mesa_set_add(def_state->uses, src);
} else {
validate_assert(state, state->if_stmt);
nir/nir: Use a linked list instead of a hash set for use/def sets This commit switches us from the current setup of using hash sets for use/def sets to using linked lists. Doing so should save us quite a bit of memory because we aren't carrying around 3 hash sets per register and 2 per SSA value. It should also save us CPU time because adding/removing things from use/def sets is 4 pointer manipulations instead of a hash lookup. Running shader-db 50 times with USE_NIR=0, NIR, and NIR + use/def lists: GLSL IR Only: 586.4 +/- 1.653833 NIR with hash sets: 675.4 +/- 2.502108 NIR + use/def lists: 641.2 +/- 1.557043 I also ran a memory usage experiment with Ken's patch to delete GLSL IR and keep NIR. This patch cuts an aditional 42.9 MiB of ralloc'd memory over and above what we gained by deleting the GLSL IR on the same dota trace. On the code complexity side of things, some things are now much easier and others are a bit harder. One of the operations we perform constantly in optimization passes is to replace one source with another. Due to the fact that an instruction can use the same SSA value multiple times, we had to iterate through the sources of the instruction and determine if the use we were replacing was the only one before removing it from the set of uses. With this patch, uses are per-source not per-instruction so we can just remove it safely. On the other hand, trying to iterate over all of the instructions that use a given value is more difficult. Fortunately, the two places we do that are the ffma peephole where it doesn't matter and GCM where we already gracefully handle duplicates visits to an instruction. Another aspect here is that using linked lists in this way can be tricky to get right. With sets, things were quite forgiving and the worst that happened if you didn't properly remove a use was that it would get caught in the validator. With linked lists, it can lead to linked list corruption which can be harder to track. However, we do just as much validation of the linked lists as we did of the sets so the validator should still catch these problems. While working on this series, the vast majority of the bugs I had to fix were caught by assertions. I don't think the lists are going to be that much worse than the sets. Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
2015-04-24 10:16:27 -07:00
_mesa_set_add(def_state->if_uses, src);
}
if (bit_size)
validate_assert(state, src->ssa->bit_size == bit_size);
if (num_components)
validate_assert(state, src->ssa->num_components == num_components);
/* TODO validate that the use is dominated by the definition */
}
static void
validate_src(nir_src *src, validate_state *state,
unsigned bit_size, unsigned num_components)
{
nir/nir: Use a linked list instead of a hash set for use/def sets This commit switches us from the current setup of using hash sets for use/def sets to using linked lists. Doing so should save us quite a bit of memory because we aren't carrying around 3 hash sets per register and 2 per SSA value. It should also save us CPU time because adding/removing things from use/def sets is 4 pointer manipulations instead of a hash lookup. Running shader-db 50 times with USE_NIR=0, NIR, and NIR + use/def lists: GLSL IR Only: 586.4 +/- 1.653833 NIR with hash sets: 675.4 +/- 2.502108 NIR + use/def lists: 641.2 +/- 1.557043 I also ran a memory usage experiment with Ken's patch to delete GLSL IR and keep NIR. This patch cuts an aditional 42.9 MiB of ralloc'd memory over and above what we gained by deleting the GLSL IR on the same dota trace. On the code complexity side of things, some things are now much easier and others are a bit harder. One of the operations we perform constantly in optimization passes is to replace one source with another. Due to the fact that an instruction can use the same SSA value multiple times, we had to iterate through the sources of the instruction and determine if the use we were replacing was the only one before removing it from the set of uses. With this patch, uses are per-source not per-instruction so we can just remove it safely. On the other hand, trying to iterate over all of the instructions that use a given value is more difficult. Fortunately, the two places we do that are the ffma peephole where it doesn't matter and GCM where we already gracefully handle duplicates visits to an instruction. Another aspect here is that using linked lists in this way can be tricky to get right. With sets, things were quite forgiving and the worst that happened if you didn't properly remove a use was that it would get caught in the validator. With linked lists, it can lead to linked list corruption which can be harder to track. However, we do just as much validation of the linked lists as we did of the sets so the validator should still catch these problems. While working on this series, the vast majority of the bugs I had to fix were caught by assertions. I don't think the lists are going to be that much worse than the sets. Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
2015-04-24 10:16:27 -07:00
if (state->instr)
validate_assert(state, src->parent_instr == state->instr);
nir/nir: Use a linked list instead of a hash set for use/def sets This commit switches us from the current setup of using hash sets for use/def sets to using linked lists. Doing so should save us quite a bit of memory because we aren't carrying around 3 hash sets per register and 2 per SSA value. It should also save us CPU time because adding/removing things from use/def sets is 4 pointer manipulations instead of a hash lookup. Running shader-db 50 times with USE_NIR=0, NIR, and NIR + use/def lists: GLSL IR Only: 586.4 +/- 1.653833 NIR with hash sets: 675.4 +/- 2.502108 NIR + use/def lists: 641.2 +/- 1.557043 I also ran a memory usage experiment with Ken's patch to delete GLSL IR and keep NIR. This patch cuts an aditional 42.9 MiB of ralloc'd memory over and above what we gained by deleting the GLSL IR on the same dota trace. On the code complexity side of things, some things are now much easier and others are a bit harder. One of the operations we perform constantly in optimization passes is to replace one source with another. Due to the fact that an instruction can use the same SSA value multiple times, we had to iterate through the sources of the instruction and determine if the use we were replacing was the only one before removing it from the set of uses. With this patch, uses are per-source not per-instruction so we can just remove it safely. On the other hand, trying to iterate over all of the instructions that use a given value is more difficult. Fortunately, the two places we do that are the ffma peephole where it doesn't matter and GCM where we already gracefully handle duplicates visits to an instruction. Another aspect here is that using linked lists in this way can be tricky to get right. With sets, things were quite forgiving and the worst that happened if you didn't properly remove a use was that it would get caught in the validator. With linked lists, it can lead to linked list corruption which can be harder to track. However, we do just as much validation of the linked lists as we did of the sets so the validator should still catch these problems. While working on this series, the vast majority of the bugs I had to fix were caught by assertions. I don't think the lists are going to be that much worse than the sets. Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
2015-04-24 10:16:27 -07:00
else
validate_assert(state, src->parent_if == state->if_stmt);
nir/nir: Use a linked list instead of a hash set for use/def sets This commit switches us from the current setup of using hash sets for use/def sets to using linked lists. Doing so should save us quite a bit of memory because we aren't carrying around 3 hash sets per register and 2 per SSA value. It should also save us CPU time because adding/removing things from use/def sets is 4 pointer manipulations instead of a hash lookup. Running shader-db 50 times with USE_NIR=0, NIR, and NIR + use/def lists: GLSL IR Only: 586.4 +/- 1.653833 NIR with hash sets: 675.4 +/- 2.502108 NIR + use/def lists: 641.2 +/- 1.557043 I also ran a memory usage experiment with Ken's patch to delete GLSL IR and keep NIR. This patch cuts an aditional 42.9 MiB of ralloc'd memory over and above what we gained by deleting the GLSL IR on the same dota trace. On the code complexity side of things, some things are now much easier and others are a bit harder. One of the operations we perform constantly in optimization passes is to replace one source with another. Due to the fact that an instruction can use the same SSA value multiple times, we had to iterate through the sources of the instruction and determine if the use we were replacing was the only one before removing it from the set of uses. With this patch, uses are per-source not per-instruction so we can just remove it safely. On the other hand, trying to iterate over all of the instructions that use a given value is more difficult. Fortunately, the two places we do that are the ffma peephole where it doesn't matter and GCM where we already gracefully handle duplicates visits to an instruction. Another aspect here is that using linked lists in this way can be tricky to get right. With sets, things were quite forgiving and the worst that happened if you didn't properly remove a use was that it would get caught in the validator. With linked lists, it can lead to linked list corruption which can be harder to track. However, we do just as much validation of the linked lists as we did of the sets so the validator should still catch these problems. While working on this series, the vast majority of the bugs I had to fix were caught by assertions. I don't think the lists are going to be that much worse than the sets. Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
2015-04-24 10:16:27 -07:00
if (src->is_ssa)
validate_ssa_src(src, state, bit_size, num_components);
else
validate_reg_src(src, state, bit_size, num_components);
}
static void
validate_alu_src(nir_alu_instr *instr, unsigned index, validate_state *state)
{
nir_alu_src *src = &instr->src[index];
unsigned num_components = nir_src_num_components(src->src);
if (!src->src.is_ssa && src->src.reg.reg->is_packed)
num_components = 4; /* can't check anything */
for (unsigned i = 0; i < 4; i++) {
validate_assert(state, src->swizzle[i] < 4);
if (nir_alu_instr_channel_used(instr, index, i))
validate_assert(state, src->swizzle[i] < num_components);
}
validate_src(&src->src, state, 0, 0);
}
static void
validate_reg_dest(nir_reg_dest *dest, validate_state *state,
unsigned bit_size, unsigned num_components)
{
validate_assert(state, dest->reg != NULL);
validate_assert(state, dest->parent_instr == state->instr);
struct hash_entry *entry2;
entry2 = _mesa_hash_table_search(state->regs, dest->reg);
validate_assert(state, entry2);
reg_validate_state *reg_state = (reg_validate_state *) entry2->data;
nir/nir: Use a linked list instead of a hash set for use/def sets This commit switches us from the current setup of using hash sets for use/def sets to using linked lists. Doing so should save us quite a bit of memory because we aren't carrying around 3 hash sets per register and 2 per SSA value. It should also save us CPU time because adding/removing things from use/def sets is 4 pointer manipulations instead of a hash lookup. Running shader-db 50 times with USE_NIR=0, NIR, and NIR + use/def lists: GLSL IR Only: 586.4 +/- 1.653833 NIR with hash sets: 675.4 +/- 2.502108 NIR + use/def lists: 641.2 +/- 1.557043 I also ran a memory usage experiment with Ken's patch to delete GLSL IR and keep NIR. This patch cuts an aditional 42.9 MiB of ralloc'd memory over and above what we gained by deleting the GLSL IR on the same dota trace. On the code complexity side of things, some things are now much easier and others are a bit harder. One of the operations we perform constantly in optimization passes is to replace one source with another. Due to the fact that an instruction can use the same SSA value multiple times, we had to iterate through the sources of the instruction and determine if the use we were replacing was the only one before removing it from the set of uses. With this patch, uses are per-source not per-instruction so we can just remove it safely. On the other hand, trying to iterate over all of the instructions that use a given value is more difficult. Fortunately, the two places we do that are the ffma peephole where it doesn't matter and GCM where we already gracefully handle duplicates visits to an instruction. Another aspect here is that using linked lists in this way can be tricky to get right. With sets, things were quite forgiving and the worst that happened if you didn't properly remove a use was that it would get caught in the validator. With linked lists, it can lead to linked list corruption which can be harder to track. However, we do just as much validation of the linked lists as we did of the sets so the validator should still catch these problems. While working on this series, the vast majority of the bugs I had to fix were caught by assertions. I don't think the lists are going to be that much worse than the sets. Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
2015-04-24 10:16:27 -07:00
_mesa_set_add(reg_state->defs, dest);
if (!dest->reg->is_global) {
validate_assert(state, reg_state->where_defined == state->impl &&
"writing to a register declared in a different function");
}
if (!dest->reg->is_packed) {
if (bit_size)
validate_assert(state, dest->reg->bit_size == bit_size);
if (num_components)
validate_assert(state, dest->reg->num_components == num_components);
}
validate_assert(state, (dest->reg->num_array_elems == 0 ||
dest->base_offset < dest->reg->num_array_elems) &&
"definitely out-of-bounds array access");
if (dest->indirect) {
validate_assert(state, dest->reg->num_array_elems != 0);
validate_assert(state, (dest->indirect->is_ssa || dest->indirect->reg.indirect == NULL) &&
"only one level of indirection allowed");
validate_src(dest->indirect, state, 32, 1);
}
}
static void
validate_ssa_def(nir_ssa_def *def, validate_state *state)
{
validate_assert(state, def->index < state->impl->ssa_alloc);
validate_assert(state, !BITSET_TEST(state->ssa_defs_found, def->index));
BITSET_SET(state->ssa_defs_found, def->index);
validate_assert(state, def->parent_instr == state->instr);
validate_assert(state, (def->num_components <= 4) ||
(def->num_components == 8) ||
(def->num_components == 16));
nir/nir: Use a linked list instead of a hash set for use/def sets This commit switches us from the current setup of using hash sets for use/def sets to using linked lists. Doing so should save us quite a bit of memory because we aren't carrying around 3 hash sets per register and 2 per SSA value. It should also save us CPU time because adding/removing things from use/def sets is 4 pointer manipulations instead of a hash lookup. Running shader-db 50 times with USE_NIR=0, NIR, and NIR + use/def lists: GLSL IR Only: 586.4 +/- 1.653833 NIR with hash sets: 675.4 +/- 2.502108 NIR + use/def lists: 641.2 +/- 1.557043 I also ran a memory usage experiment with Ken's patch to delete GLSL IR and keep NIR. This patch cuts an aditional 42.9 MiB of ralloc'd memory over and above what we gained by deleting the GLSL IR on the same dota trace. On the code complexity side of things, some things are now much easier and others are a bit harder. One of the operations we perform constantly in optimization passes is to replace one source with another. Due to the fact that an instruction can use the same SSA value multiple times, we had to iterate through the sources of the instruction and determine if the use we were replacing was the only one before removing it from the set of uses. With this patch, uses are per-source not per-instruction so we can just remove it safely. On the other hand, trying to iterate over all of the instructions that use a given value is more difficult. Fortunately, the two places we do that are the ffma peephole where it doesn't matter and GCM where we already gracefully handle duplicates visits to an instruction. Another aspect here is that using linked lists in this way can be tricky to get right. With sets, things were quite forgiving and the worst that happened if you didn't properly remove a use was that it would get caught in the validator. With linked lists, it can lead to linked list corruption which can be harder to track. However, we do just as much validation of the linked lists as we did of the sets so the validator should still catch these problems. While working on this series, the vast majority of the bugs I had to fix were caught by assertions. I don't think the lists are going to be that much worse than the sets. Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
2015-04-24 10:16:27 -07:00
list_validate(&def->uses);
list_validate(&def->if_uses);
ssa_def_validate_state *def_state = ralloc(state->ssa_defs,
ssa_def_validate_state);
def_state->where_defined = state->impl;
def_state->uses = _mesa_set_create(def_state, _mesa_hash_pointer,
_mesa_key_pointer_equal);
def_state->if_uses = _mesa_set_create(def_state, _mesa_hash_pointer,
_mesa_key_pointer_equal);
_mesa_hash_table_insert(state->ssa_defs, def, def_state);
}
static void
validate_dest(nir_dest *dest, validate_state *state,
unsigned bit_size, unsigned num_components)
{
if (dest->is_ssa) {
if (bit_size)
validate_assert(state, dest->ssa.bit_size == bit_size);
if (num_components)
validate_assert(state, dest->ssa.num_components == num_components);
validate_ssa_def(&dest->ssa, state);
} else {
validate_reg_dest(&dest->reg, state, bit_size, num_components);
}
}
static void
validate_alu_dest(nir_alu_instr *instr, validate_state *state)
{
nir_alu_dest *dest = &instr->dest;
unsigned dest_size = nir_dest_num_components(dest->dest);
bool is_packed = !dest->dest.is_ssa && dest->dest.reg.reg->is_packed;
/*
* validate that the instruction doesn't write to components not in the
* register/SSA value
*/
validate_assert(state, is_packed || !(dest->write_mask & ~((1 << dest_size) - 1)));
/* validate that saturate is only ever used on instructions with
* destinations of type float
*/
nir_alu_instr *alu = nir_instr_as_alu(state->instr);
validate_assert(state,
(nir_alu_type_get_base_type(nir_op_infos[alu->op].output_type) ==
nir_type_float) ||
!dest->saturate);
validate_dest(&dest->dest, state, 0, 0);
}
static void
validate_alu_instr(nir_alu_instr *instr, validate_state *state)
{
validate_assert(state, instr->op < nir_num_opcodes);
unsigned instr_bit_size = 0;
for (unsigned i = 0; i < nir_op_infos[instr->op].num_inputs; i++) {
nir_alu_type src_type = nir_op_infos[instr->op].input_types[i];
unsigned src_bit_size = nir_src_bit_size(instr->src[i].src);
if (nir_alu_type_get_type_size(src_type)) {
validate_assert(state, src_bit_size == nir_alu_type_get_type_size(src_type));
} else if (instr_bit_size) {
validate_assert(state, src_bit_size == instr_bit_size);
} else {
instr_bit_size = src_bit_size;
}
if (nir_alu_type_get_base_type(src_type) == nir_type_float) {
/* 8-bit float isn't a thing */
validate_assert(state, src_bit_size == 16 || src_bit_size == 32 ||
src_bit_size == 64);
}
validate_alu_src(instr, i, state);
}
nir_alu_type dest_type = nir_op_infos[instr->op].output_type;
unsigned dest_bit_size = nir_dest_bit_size(instr->dest.dest);
if (nir_alu_type_get_type_size(dest_type)) {
validate_assert(state, dest_bit_size == nir_alu_type_get_type_size(dest_type));
} else if (instr_bit_size) {
validate_assert(state, dest_bit_size == instr_bit_size);
} else {
/* The only unsized thing is the destination so it's vacuously valid */
}
if (nir_alu_type_get_base_type(dest_type) == nir_type_float) {
/* 8-bit float isn't a thing */
validate_assert(state, dest_bit_size == 16 || dest_bit_size == 32 ||
dest_bit_size == 64);
}
validate_alu_dest(instr, state);
}
static void
validate_var_use(nir_variable *var, validate_state *state)
{
struct hash_entry *entry = _mesa_hash_table_search(state->var_defs, var);
validate_assert(state, entry);
if (var->data.mode == nir_var_local)
validate_assert(state, (nir_function_impl *) entry->data == state->impl);
}
static void
validate_deref_instr(nir_deref_instr *instr, validate_state *state)
{
if (instr->deref_type == nir_deref_type_var) {
/* Variable dereferences are stupid simple. */
validate_assert(state, instr->mode == instr->var->data.mode);
validate_assert(state, instr->type == instr->var->type);
validate_var_use(instr->var, state);
} else if (instr->deref_type == nir_deref_type_cast) {
/* For cast, we simply have to trust the instruction. It's up to
* lowering passes and front/back-ends to make them sane.
*/
validate_src(&instr->parent, state, 0, 0);
/* We just validate that the type and mode are there */
validate_assert(state, instr->mode);
validate_assert(state, instr->type);
} else {
/* We require the parent to be SSA. This may be lifted in the future */
validate_assert(state, instr->parent.is_ssa);
/* The parent pointer value must have the same number of components
* as the destination.
*/
validate_src(&instr->parent, state, nir_dest_bit_size(instr->dest),
nir_dest_num_components(instr->dest));
nir_instr *parent_instr = instr->parent.ssa->parent_instr;
/* The parent must come from another deref instruction */
validate_assert(state, parent_instr->type == nir_instr_type_deref);
nir_deref_instr *parent = nir_instr_as_deref(parent_instr);
validate_assert(state, instr->mode == parent->mode);
switch (instr->deref_type) {
case nir_deref_type_struct:
validate_assert(state, glsl_type_is_struct(parent->type));
validate_assert(state,
instr->strct.index < glsl_get_length(parent->type));
validate_assert(state, instr->type ==
glsl_get_struct_field(parent->type, instr->strct.index));
break;
case nir_deref_type_array:
case nir_deref_type_array_wildcard:
if (instr->mode == nir_var_shared) {
/* Shared variables have a bit more relaxed rules because we need
* to be able to handle array derefs on vectors. Fortunately,
* nir_lower_io handles these just fine.
*/
validate_assert(state, glsl_type_is_array(parent->type) ||
glsl_type_is_matrix(parent->type) ||
glsl_type_is_vector(parent->type));
} else {
/* Most of NIR cannot handle array derefs on vectors */
validate_assert(state, glsl_type_is_array(parent->type) ||
glsl_type_is_matrix(parent->type));
}
validate_assert(state,
instr->type == glsl_get_array_element(parent->type));
if (instr->deref_type == nir_deref_type_array)
validate_src(&instr->arr.index, state, 32, 1);
break;
default:
unreachable("Invalid deref instruction type");
}
}
/* We intentionally don't validate the size of the destination because we
* want to let other compiler components such as SPIR-V decide how big
* pointers should be.
*/
validate_dest(&instr->dest, state, 0, 0);
}
static void
validate_intrinsic_instr(nir_intrinsic_instr *instr, validate_state *state)
{
unsigned dest_bit_size = 0;
unsigned src_bit_sizes[NIR_INTRINSIC_MAX_INPUTS] = { 0, };
switch (instr->intrinsic) {
case nir_intrinsic_load_param: {
unsigned param_idx = nir_intrinsic_param_idx(instr);
validate_assert(state, param_idx < state->impl->function->num_params);
nir_parameter *param = &state->impl->function->params[param_idx];
validate_assert(state, instr->num_components == param->num_components);
dest_bit_size = param->bit_size;
break;
}
case nir_intrinsic_load_deref: {
nir_deref_instr *src = nir_src_as_deref(instr->src[0]);
validate_assert(state, glsl_type_is_vector_or_scalar(src->type) ||
(src->mode == nir_var_uniform &&
glsl_get_base_type(src->type) == GLSL_TYPE_SUBROUTINE));
validate_assert(state, instr->num_components ==
glsl_get_vector_elements(src->type));
dest_bit_size = glsl_get_bit_size(src->type);
break;
}
case nir_intrinsic_store_deref: {
nir_deref_instr *dst = nir_src_as_deref(instr->src[0]);
validate_assert(state, glsl_type_is_vector_or_scalar(dst->type));
validate_assert(state, instr->num_components ==
glsl_get_vector_elements(dst->type));
src_bit_sizes[1] = glsl_get_bit_size(dst->type);
validate_assert(state, (dst->mode & (nir_var_shader_in |
nir_var_uniform |
nir_var_shader_storage)) == 0);
validate_assert(state, (nir_intrinsic_write_mask(instr) & ~((1 << instr->num_components) - 1)) == 0);
break;
}
case nir_intrinsic_copy_deref: {
nir_deref_instr *dst = nir_src_as_deref(instr->src[0]);
nir_deref_instr *src = nir_src_as_deref(instr->src[1]);
validate_assert(state, dst->type == src->type);
validate_assert(state, (dst->mode & (nir_var_shader_in |
nir_var_uniform |
nir_var_shader_storage)) == 0);
break;
}
default:
break;
}
unsigned num_srcs = nir_intrinsic_infos[instr->intrinsic].num_srcs;
for (unsigned i = 0; i < num_srcs; i++) {
unsigned components_read = nir_intrinsic_src_components(instr, i);
validate_assert(state, components_read > 0);
validate_src(&instr->src[i], state, src_bit_sizes[i], components_read);
}
if (nir_intrinsic_infos[instr->intrinsic].has_dest) {
unsigned components_written = nir_intrinsic_dest_components(instr);
validate_assert(state, components_written > 0);
validate_dest(&instr->dest, state, dest_bit_size, components_written);
}
}
static void
validate_tex_instr(nir_tex_instr *instr, validate_state *state)
{
bool src_type_seen[nir_num_tex_src_types];
for (unsigned i = 0; i < nir_num_tex_src_types; i++)
src_type_seen[i] = false;
for (unsigned i = 0; i < instr->num_srcs; i++) {
validate_assert(state, !src_type_seen[instr->src[i].src_type]);
src_type_seen[instr->src[i].src_type] = true;
validate_src(&instr->src[i].src, state,
0, nir_tex_instr_src_size(instr, i));
}
validate_dest(&instr->dest, state, 0, nir_tex_instr_dest_size(instr));
}
static void
validate_call_instr(nir_call_instr *instr, validate_state *state)
{
validate_assert(state, instr->num_params == instr->callee->num_params);
for (unsigned i = 0; i < instr->num_params; i++) {
validate_src(&instr->params[i], state,
instr->callee->params[i].bit_size,
instr->callee->params[i].num_components);
}
}
static void
validate_load_const_instr(nir_load_const_instr *instr, validate_state *state)
{
validate_ssa_def(&instr->def, state);
}
static void
validate_ssa_undef_instr(nir_ssa_undef_instr *instr, validate_state *state)
{
validate_ssa_def(&instr->def, state);
}
static void
validate_phi_instr(nir_phi_instr *instr, validate_state *state)
{
/*
* don't validate the sources until we get to them from their predecessor
* basic blocks, to avoid validating an SSA use before its definition.
*/
validate_dest(&instr->dest, state, 0, 0);
exec_list_validate(&instr->srcs);
validate_assert(state, exec_list_length(&instr->srcs) ==
state->block->predecessors->entries);
}
static void
validate_instr(nir_instr *instr, validate_state *state)
{
validate_assert(state, instr->block == state->block);
state->instr = instr;
switch (instr->type) {
case nir_instr_type_alu:
validate_alu_instr(nir_instr_as_alu(instr), state);
break;
case nir_instr_type_deref:
validate_deref_instr(nir_instr_as_deref(instr), state);
break;
case nir_instr_type_call:
validate_call_instr(nir_instr_as_call(instr), state);
break;
case nir_instr_type_intrinsic:
validate_intrinsic_instr(nir_instr_as_intrinsic(instr), state);
break;
case nir_instr_type_tex:
validate_tex_instr(nir_instr_as_tex(instr), state);
break;
case nir_instr_type_load_const:
validate_load_const_instr(nir_instr_as_load_const(instr), state);
break;
case nir_instr_type_phi:
validate_phi_instr(nir_instr_as_phi(instr), state);
break;
case nir_instr_type_ssa_undef:
validate_ssa_undef_instr(nir_instr_as_ssa_undef(instr), state);
break;
case nir_instr_type_jump:
break;
default:
validate_assert(state, !"Invalid ALU instruction type");
break;
}
state->instr = NULL;
}
static void
validate_phi_src(nir_phi_instr *instr, nir_block *pred, validate_state *state)
{
state->instr = &instr->instr;
validate_assert(state, instr->dest.is_ssa);
exec_list_validate(&instr->srcs);
nir_foreach_phi_src(src, instr) {
if (src->pred == pred) {
validate_assert(state, src->src.is_ssa);
validate_src(&src->src, state, instr->dest.ssa.bit_size,
instr->dest.ssa.num_components);
state->instr = NULL;
return;
}
}
abort();
}
static void
validate_phi_srcs(nir_block *block, nir_block *succ, validate_state *state)
{
nir_foreach_instr(instr, succ) {
if (instr->type != nir_instr_type_phi)
break;
validate_phi_src(nir_instr_as_phi(instr), block, state);
}
}
static void validate_cf_node(nir_cf_node *node, validate_state *state);
static void
validate_block(nir_block *block, validate_state *state)
{
validate_assert(state, block->cf_node.parent == state->parent_node);
state->block = block;
exec_list_validate(&block->instr_list);
nir_foreach_instr(instr, block) {
if (instr->type == nir_instr_type_phi) {
validate_assert(state, instr == nir_block_first_instr(block) ||
nir_instr_prev(instr)->type == nir_instr_type_phi);
}
if (instr->type == nir_instr_type_jump) {
validate_assert(state, instr == nir_block_last_instr(block));
}
validate_instr(instr, state);
}
validate_assert(state, block->successors[0] != NULL);
validate_assert(state, block->successors[0] != block->successors[1]);
for (unsigned i = 0; i < 2; i++) {
if (block->successors[i] != NULL) {
struct set_entry *entry =
_mesa_set_search(block->successors[i]->predecessors, block);
validate_assert(state, entry);
validate_phi_srcs(block, block->successors[i], state);
}
}
struct set_entry *entry;
set_foreach(block->predecessors, entry) {
const nir_block *pred = entry->key;
validate_assert(state, pred->successors[0] == block ||
pred->successors[1] == block);
}
if (!exec_list_is_empty(&block->instr_list) &&
nir_block_last_instr(block)->type == nir_instr_type_jump) {
validate_assert(state, block->successors[1] == NULL);
nir_jump_instr *jump = nir_instr_as_jump(nir_block_last_instr(block));
switch (jump->type) {
case nir_jump_break: {
nir_block *after =
nir_cf_node_as_block(nir_cf_node_next(&state->loop->cf_node));
validate_assert(state, block->successors[0] == after);
break;
}
case nir_jump_continue: {
nir_block *first = nir_loop_first_block(state->loop);
validate_assert(state, block->successors[0] == first);
break;
}
case nir_jump_return:
validate_assert(state, block->successors[0] == state->impl->end_block);
break;
default:
unreachable("bad jump type");
}
} else {
nir_cf_node *next = nir_cf_node_next(&block->cf_node);
if (next == NULL) {
switch (state->parent_node->type) {
case nir_cf_node_loop: {
nir_block *first = nir_loop_first_block(state->loop);
validate_assert(state, block->successors[0] == first);
/* due to the hack for infinite loops, block->successors[1] may
* point to the block after the loop.
*/
break;
}
case nir_cf_node_if: {
nir_block *after =
nir_cf_node_as_block(nir_cf_node_next(state->parent_node));
validate_assert(state, block->successors[0] == after);
validate_assert(state, block->successors[1] == NULL);
break;
}
case nir_cf_node_function:
validate_assert(state, block->successors[0] == state->impl->end_block);
validate_assert(state, block->successors[1] == NULL);
break;
default:
unreachable("unknown control flow node type");
}
} else {
if (next->type == nir_cf_node_if) {
nir_if *if_stmt = nir_cf_node_as_if(next);
validate_assert(state, block->successors[0] ==
nir_if_first_then_block(if_stmt));
validate_assert(state, block->successors[1] ==
nir_if_first_else_block(if_stmt));
} else {
validate_assert(state, next->type == nir_cf_node_loop);
nir_loop *loop = nir_cf_node_as_loop(next);
validate_assert(state, block->successors[0] ==
nir_loop_first_block(loop));
validate_assert(state, block->successors[1] == NULL);
}
}
}
}
static void
validate_if(nir_if *if_stmt, validate_state *state)
{
state->if_stmt = if_stmt;
validate_assert(state, !exec_node_is_head_sentinel(if_stmt->cf_node.node.prev));
nir_cf_node *prev_node = nir_cf_node_prev(&if_stmt->cf_node);
validate_assert(state, prev_node->type == nir_cf_node_block);
validate_assert(state, !exec_node_is_tail_sentinel(if_stmt->cf_node.node.next));
nir_cf_node *next_node = nir_cf_node_next(&if_stmt->cf_node);
validate_assert(state, next_node->type == nir_cf_node_block);
validate_src(&if_stmt->condition, state, 32, 1);
validate_assert(state, !exec_list_is_empty(&if_stmt->then_list));
validate_assert(state, !exec_list_is_empty(&if_stmt->else_list));
nir_cf_node *old_parent = state->parent_node;
state->parent_node = &if_stmt->cf_node;
exec_list_validate(&if_stmt->then_list);
foreach_list_typed(nir_cf_node, cf_node, node, &if_stmt->then_list) {
validate_cf_node(cf_node, state);
}
exec_list_validate(&if_stmt->else_list);
foreach_list_typed(nir_cf_node, cf_node, node, &if_stmt->else_list) {
validate_cf_node(cf_node, state);
}
state->parent_node = old_parent;
state->if_stmt = NULL;
}
static void
validate_loop(nir_loop *loop, validate_state *state)
{
validate_assert(state, !exec_node_is_head_sentinel(loop->cf_node.node.prev));
nir_cf_node *prev_node = nir_cf_node_prev(&loop->cf_node);
validate_assert(state, prev_node->type == nir_cf_node_block);
validate_assert(state, !exec_node_is_tail_sentinel(loop->cf_node.node.next));
nir_cf_node *next_node = nir_cf_node_next(&loop->cf_node);
validate_assert(state, next_node->type == nir_cf_node_block);
validate_assert(state, !exec_list_is_empty(&loop->body));
nir_cf_node *old_parent = state->parent_node;
state->parent_node = &loop->cf_node;
nir_loop *old_loop = state->loop;
state->loop = loop;
exec_list_validate(&loop->body);
foreach_list_typed(nir_cf_node, cf_node, node, &loop->body) {
validate_cf_node(cf_node, state);
}
state->parent_node = old_parent;
state->loop = old_loop;
}
static void
validate_cf_node(nir_cf_node *node, validate_state *state)
{
validate_assert(state, node->parent == state->parent_node);
switch (node->type) {
case nir_cf_node_block:
validate_block(nir_cf_node_as_block(node), state);
break;
case nir_cf_node_if:
validate_if(nir_cf_node_as_if(node), state);
break;
case nir_cf_node_loop:
validate_loop(nir_cf_node_as_loop(node), state);
break;
default:
unreachable("Invalid CF node type");
}
}
static void
prevalidate_reg_decl(nir_register *reg, bool is_global, validate_state *state)
{
validate_assert(state, reg->is_global == is_global);
if (is_global)
validate_assert(state, reg->index < state->shader->reg_alloc);
else
validate_assert(state, reg->index < state->impl->reg_alloc);
validate_assert(state, !BITSET_TEST(state->regs_found, reg->index));
BITSET_SET(state->regs_found, reg->index);
nir/nir: Use a linked list instead of a hash set for use/def sets This commit switches us from the current setup of using hash sets for use/def sets to using linked lists. Doing so should save us quite a bit of memory because we aren't carrying around 3 hash sets per register and 2 per SSA value. It should also save us CPU time because adding/removing things from use/def sets is 4 pointer manipulations instead of a hash lookup. Running shader-db 50 times with USE_NIR=0, NIR, and NIR + use/def lists: GLSL IR Only: 586.4 +/- 1.653833 NIR with hash sets: 675.4 +/- 2.502108 NIR + use/def lists: 641.2 +/- 1.557043 I also ran a memory usage experiment with Ken's patch to delete GLSL IR and keep NIR. This patch cuts an aditional 42.9 MiB of ralloc'd memory over and above what we gained by deleting the GLSL IR on the same dota trace. On the code complexity side of things, some things are now much easier and others are a bit harder. One of the operations we perform constantly in optimization passes is to replace one source with another. Due to the fact that an instruction can use the same SSA value multiple times, we had to iterate through the sources of the instruction and determine if the use we were replacing was the only one before removing it from the set of uses. With this patch, uses are per-source not per-instruction so we can just remove it safely. On the other hand, trying to iterate over all of the instructions that use a given value is more difficult. Fortunately, the two places we do that are the ffma peephole where it doesn't matter and GCM where we already gracefully handle duplicates visits to an instruction. Another aspect here is that using linked lists in this way can be tricky to get right. With sets, things were quite forgiving and the worst that happened if you didn't properly remove a use was that it would get caught in the validator. With linked lists, it can lead to linked list corruption which can be harder to track. However, we do just as much validation of the linked lists as we did of the sets so the validator should still catch these problems. While working on this series, the vast majority of the bugs I had to fix were caught by assertions. I don't think the lists are going to be that much worse than the sets. Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
2015-04-24 10:16:27 -07:00
list_validate(&reg->uses);
list_validate(&reg->defs);
list_validate(&reg->if_uses);
reg_validate_state *reg_state = ralloc(state->regs, reg_validate_state);
reg_state->uses = _mesa_set_create(reg_state, _mesa_hash_pointer,
_mesa_key_pointer_equal);
reg_state->if_uses = _mesa_set_create(reg_state, _mesa_hash_pointer,
_mesa_key_pointer_equal);
reg_state->defs = _mesa_set_create(reg_state, _mesa_hash_pointer,
_mesa_key_pointer_equal);
reg_state->where_defined = is_global ? NULL : state->impl;
_mesa_hash_table_insert(state->regs, reg, reg_state);
}
static void
postvalidate_reg_decl(nir_register *reg, validate_state *state)
{
struct hash_entry *entry = _mesa_hash_table_search(state->regs, reg);
assume(entry);
reg_validate_state *reg_state = (reg_validate_state *) entry->data;
nir_foreach_use(src, reg) {
nir/nir: Use a linked list instead of a hash set for use/def sets This commit switches us from the current setup of using hash sets for use/def sets to using linked lists. Doing so should save us quite a bit of memory because we aren't carrying around 3 hash sets per register and 2 per SSA value. It should also save us CPU time because adding/removing things from use/def sets is 4 pointer manipulations instead of a hash lookup. Running shader-db 50 times with USE_NIR=0, NIR, and NIR + use/def lists: GLSL IR Only: 586.4 +/- 1.653833 NIR with hash sets: 675.4 +/- 2.502108 NIR + use/def lists: 641.2 +/- 1.557043 I also ran a memory usage experiment with Ken's patch to delete GLSL IR and keep NIR. This patch cuts an aditional 42.9 MiB of ralloc'd memory over and above what we gained by deleting the GLSL IR on the same dota trace. On the code complexity side of things, some things are now much easier and others are a bit harder. One of the operations we perform constantly in optimization passes is to replace one source with another. Due to the fact that an instruction can use the same SSA value multiple times, we had to iterate through the sources of the instruction and determine if the use we were replacing was the only one before removing it from the set of uses. With this patch, uses are per-source not per-instruction so we can just remove it safely. On the other hand, trying to iterate over all of the instructions that use a given value is more difficult. Fortunately, the two places we do that are the ffma peephole where it doesn't matter and GCM where we already gracefully handle duplicates visits to an instruction. Another aspect here is that using linked lists in this way can be tricky to get right. With sets, things were quite forgiving and the worst that happened if you didn't properly remove a use was that it would get caught in the validator. With linked lists, it can lead to linked list corruption which can be harder to track. However, we do just as much validation of the linked lists as we did of the sets so the validator should still catch these problems. While working on this series, the vast majority of the bugs I had to fix were caught by assertions. I don't think the lists are going to be that much worse than the sets. Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
2015-04-24 10:16:27 -07:00
struct set_entry *entry = _mesa_set_search(reg_state->uses, src);
validate_assert(state, entry);
nir/nir: Use a linked list instead of a hash set for use/def sets This commit switches us from the current setup of using hash sets for use/def sets to using linked lists. Doing so should save us quite a bit of memory because we aren't carrying around 3 hash sets per register and 2 per SSA value. It should also save us CPU time because adding/removing things from use/def sets is 4 pointer manipulations instead of a hash lookup. Running shader-db 50 times with USE_NIR=0, NIR, and NIR + use/def lists: GLSL IR Only: 586.4 +/- 1.653833 NIR with hash sets: 675.4 +/- 2.502108 NIR + use/def lists: 641.2 +/- 1.557043 I also ran a memory usage experiment with Ken's patch to delete GLSL IR and keep NIR. This patch cuts an aditional 42.9 MiB of ralloc'd memory over and above what we gained by deleting the GLSL IR on the same dota trace. On the code complexity side of things, some things are now much easier and others are a bit harder. One of the operations we perform constantly in optimization passes is to replace one source with another. Due to the fact that an instruction can use the same SSA value multiple times, we had to iterate through the sources of the instruction and determine if the use we were replacing was the only one before removing it from the set of uses. With this patch, uses are per-source not per-instruction so we can just remove it safely. On the other hand, trying to iterate over all of the instructions that use a given value is more difficult. Fortunately, the two places we do that are the ffma peephole where it doesn't matter and GCM where we already gracefully handle duplicates visits to an instruction. Another aspect here is that using linked lists in this way can be tricky to get right. With sets, things were quite forgiving and the worst that happened if you didn't properly remove a use was that it would get caught in the validator. With linked lists, it can lead to linked list corruption which can be harder to track. However, we do just as much validation of the linked lists as we did of the sets so the validator should still catch these problems. While working on this series, the vast majority of the bugs I had to fix were caught by assertions. I don't think the lists are going to be that much worse than the sets. Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
2015-04-24 10:16:27 -07:00
_mesa_set_remove(reg_state->uses, entry);
}
if (reg_state->uses->entries != 0) {
printf("extra entries in register uses:\n");
struct set_entry *entry;
nir/nir: Use a linked list instead of a hash set for use/def sets This commit switches us from the current setup of using hash sets for use/def sets to using linked lists. Doing so should save us quite a bit of memory because we aren't carrying around 3 hash sets per register and 2 per SSA value. It should also save us CPU time because adding/removing things from use/def sets is 4 pointer manipulations instead of a hash lookup. Running shader-db 50 times with USE_NIR=0, NIR, and NIR + use/def lists: GLSL IR Only: 586.4 +/- 1.653833 NIR with hash sets: 675.4 +/- 2.502108 NIR + use/def lists: 641.2 +/- 1.557043 I also ran a memory usage experiment with Ken's patch to delete GLSL IR and keep NIR. This patch cuts an aditional 42.9 MiB of ralloc'd memory over and above what we gained by deleting the GLSL IR on the same dota trace. On the code complexity side of things, some things are now much easier and others are a bit harder. One of the operations we perform constantly in optimization passes is to replace one source with another. Due to the fact that an instruction can use the same SSA value multiple times, we had to iterate through the sources of the instruction and determine if the use we were replacing was the only one before removing it from the set of uses. With this patch, uses are per-source not per-instruction so we can just remove it safely. On the other hand, trying to iterate over all of the instructions that use a given value is more difficult. Fortunately, the two places we do that are the ffma peephole where it doesn't matter and GCM where we already gracefully handle duplicates visits to an instruction. Another aspect here is that using linked lists in this way can be tricky to get right. With sets, things were quite forgiving and the worst that happened if you didn't properly remove a use was that it would get caught in the validator. With linked lists, it can lead to linked list corruption which can be harder to track. However, we do just as much validation of the linked lists as we did of the sets so the validator should still catch these problems. While working on this series, the vast majority of the bugs I had to fix were caught by assertions. I don't think the lists are going to be that much worse than the sets. Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
2015-04-24 10:16:27 -07:00
set_foreach(reg_state->uses, entry)
printf("%p\n", entry->key);
abort();
}
nir_foreach_if_use(src, reg) {
nir/nir: Use a linked list instead of a hash set for use/def sets This commit switches us from the current setup of using hash sets for use/def sets to using linked lists. Doing so should save us quite a bit of memory because we aren't carrying around 3 hash sets per register and 2 per SSA value. It should also save us CPU time because adding/removing things from use/def sets is 4 pointer manipulations instead of a hash lookup. Running shader-db 50 times with USE_NIR=0, NIR, and NIR + use/def lists: GLSL IR Only: 586.4 +/- 1.653833 NIR with hash sets: 675.4 +/- 2.502108 NIR + use/def lists: 641.2 +/- 1.557043 I also ran a memory usage experiment with Ken's patch to delete GLSL IR and keep NIR. This patch cuts an aditional 42.9 MiB of ralloc'd memory over and above what we gained by deleting the GLSL IR on the same dota trace. On the code complexity side of things, some things are now much easier and others are a bit harder. One of the operations we perform constantly in optimization passes is to replace one source with another. Due to the fact that an instruction can use the same SSA value multiple times, we had to iterate through the sources of the instruction and determine if the use we were replacing was the only one before removing it from the set of uses. With this patch, uses are per-source not per-instruction so we can just remove it safely. On the other hand, trying to iterate over all of the instructions that use a given value is more difficult. Fortunately, the two places we do that are the ffma peephole where it doesn't matter and GCM where we already gracefully handle duplicates visits to an instruction. Another aspect here is that using linked lists in this way can be tricky to get right. With sets, things were quite forgiving and the worst that happened if you didn't properly remove a use was that it would get caught in the validator. With linked lists, it can lead to linked list corruption which can be harder to track. However, we do just as much validation of the linked lists as we did of the sets so the validator should still catch these problems. While working on this series, the vast majority of the bugs I had to fix were caught by assertions. I don't think the lists are going to be that much worse than the sets. Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
2015-04-24 10:16:27 -07:00
struct set_entry *entry = _mesa_set_search(reg_state->if_uses, src);
validate_assert(state, entry);
nir/nir: Use a linked list instead of a hash set for use/def sets This commit switches us from the current setup of using hash sets for use/def sets to using linked lists. Doing so should save us quite a bit of memory because we aren't carrying around 3 hash sets per register and 2 per SSA value. It should also save us CPU time because adding/removing things from use/def sets is 4 pointer manipulations instead of a hash lookup. Running shader-db 50 times with USE_NIR=0, NIR, and NIR + use/def lists: GLSL IR Only: 586.4 +/- 1.653833 NIR with hash sets: 675.4 +/- 2.502108 NIR + use/def lists: 641.2 +/- 1.557043 I also ran a memory usage experiment with Ken's patch to delete GLSL IR and keep NIR. This patch cuts an aditional 42.9 MiB of ralloc'd memory over and above what we gained by deleting the GLSL IR on the same dota trace. On the code complexity side of things, some things are now much easier and others are a bit harder. One of the operations we perform constantly in optimization passes is to replace one source with another. Due to the fact that an instruction can use the same SSA value multiple times, we had to iterate through the sources of the instruction and determine if the use we were replacing was the only one before removing it from the set of uses. With this patch, uses are per-source not per-instruction so we can just remove it safely. On the other hand, trying to iterate over all of the instructions that use a given value is more difficult. Fortunately, the two places we do that are the ffma peephole where it doesn't matter and GCM where we already gracefully handle duplicates visits to an instruction. Another aspect here is that using linked lists in this way can be tricky to get right. With sets, things were quite forgiving and the worst that happened if you didn't properly remove a use was that it would get caught in the validator. With linked lists, it can lead to linked list corruption which can be harder to track. However, we do just as much validation of the linked lists as we did of the sets so the validator should still catch these problems. While working on this series, the vast majority of the bugs I had to fix were caught by assertions. I don't think the lists are going to be that much worse than the sets. Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
2015-04-24 10:16:27 -07:00
_mesa_set_remove(reg_state->if_uses, entry);
}
if (reg_state->if_uses->entries != 0) {
printf("extra entries in register if_uses:\n");
struct set_entry *entry;
nir/nir: Use a linked list instead of a hash set for use/def sets This commit switches us from the current setup of using hash sets for use/def sets to using linked lists. Doing so should save us quite a bit of memory because we aren't carrying around 3 hash sets per register and 2 per SSA value. It should also save us CPU time because adding/removing things from use/def sets is 4 pointer manipulations instead of a hash lookup. Running shader-db 50 times with USE_NIR=0, NIR, and NIR + use/def lists: GLSL IR Only: 586.4 +/- 1.653833 NIR with hash sets: 675.4 +/- 2.502108 NIR + use/def lists: 641.2 +/- 1.557043 I also ran a memory usage experiment with Ken's patch to delete GLSL IR and keep NIR. This patch cuts an aditional 42.9 MiB of ralloc'd memory over and above what we gained by deleting the GLSL IR on the same dota trace. On the code complexity side of things, some things are now much easier and others are a bit harder. One of the operations we perform constantly in optimization passes is to replace one source with another. Due to the fact that an instruction can use the same SSA value multiple times, we had to iterate through the sources of the instruction and determine if the use we were replacing was the only one before removing it from the set of uses. With this patch, uses are per-source not per-instruction so we can just remove it safely. On the other hand, trying to iterate over all of the instructions that use a given value is more difficult. Fortunately, the two places we do that are the ffma peephole where it doesn't matter and GCM where we already gracefully handle duplicates visits to an instruction. Another aspect here is that using linked lists in this way can be tricky to get right. With sets, things were quite forgiving and the worst that happened if you didn't properly remove a use was that it would get caught in the validator. With linked lists, it can lead to linked list corruption which can be harder to track. However, we do just as much validation of the linked lists as we did of the sets so the validator should still catch these problems. While working on this series, the vast majority of the bugs I had to fix were caught by assertions. I don't think the lists are going to be that much worse than the sets. Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
2015-04-24 10:16:27 -07:00
set_foreach(reg_state->if_uses, entry)
printf("%p\n", entry->key);
abort();
}
nir_foreach_def(src, reg) {
nir/nir: Use a linked list instead of a hash set for use/def sets This commit switches us from the current setup of using hash sets for use/def sets to using linked lists. Doing so should save us quite a bit of memory because we aren't carrying around 3 hash sets per register and 2 per SSA value. It should also save us CPU time because adding/removing things from use/def sets is 4 pointer manipulations instead of a hash lookup. Running shader-db 50 times with USE_NIR=0, NIR, and NIR + use/def lists: GLSL IR Only: 586.4 +/- 1.653833 NIR with hash sets: 675.4 +/- 2.502108 NIR + use/def lists: 641.2 +/- 1.557043 I also ran a memory usage experiment with Ken's patch to delete GLSL IR and keep NIR. This patch cuts an aditional 42.9 MiB of ralloc'd memory over and above what we gained by deleting the GLSL IR on the same dota trace. On the code complexity side of things, some things are now much easier and others are a bit harder. One of the operations we perform constantly in optimization passes is to replace one source with another. Due to the fact that an instruction can use the same SSA value multiple times, we had to iterate through the sources of the instruction and determine if the use we were replacing was the only one before removing it from the set of uses. With this patch, uses are per-source not per-instruction so we can just remove it safely. On the other hand, trying to iterate over all of the instructions that use a given value is more difficult. Fortunately, the two places we do that are the ffma peephole where it doesn't matter and GCM where we already gracefully handle duplicates visits to an instruction. Another aspect here is that using linked lists in this way can be tricky to get right. With sets, things were quite forgiving and the worst that happened if you didn't properly remove a use was that it would get caught in the validator. With linked lists, it can lead to linked list corruption which can be harder to track. However, we do just as much validation of the linked lists as we did of the sets so the validator should still catch these problems. While working on this series, the vast majority of the bugs I had to fix were caught by assertions. I don't think the lists are going to be that much worse than the sets. Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
2015-04-24 10:16:27 -07:00
struct set_entry *entry = _mesa_set_search(reg_state->defs, src);
validate_assert(state, entry);
nir/nir: Use a linked list instead of a hash set for use/def sets This commit switches us from the current setup of using hash sets for use/def sets to using linked lists. Doing so should save us quite a bit of memory because we aren't carrying around 3 hash sets per register and 2 per SSA value. It should also save us CPU time because adding/removing things from use/def sets is 4 pointer manipulations instead of a hash lookup. Running shader-db 50 times with USE_NIR=0, NIR, and NIR + use/def lists: GLSL IR Only: 586.4 +/- 1.653833 NIR with hash sets: 675.4 +/- 2.502108 NIR + use/def lists: 641.2 +/- 1.557043 I also ran a memory usage experiment with Ken's patch to delete GLSL IR and keep NIR. This patch cuts an aditional 42.9 MiB of ralloc'd memory over and above what we gained by deleting the GLSL IR on the same dota trace. On the code complexity side of things, some things are now much easier and others are a bit harder. One of the operations we perform constantly in optimization passes is to replace one source with another. Due to the fact that an instruction can use the same SSA value multiple times, we had to iterate through the sources of the instruction and determine if the use we were replacing was the only one before removing it from the set of uses. With this patch, uses are per-source not per-instruction so we can just remove it safely. On the other hand, trying to iterate over all of the instructions that use a given value is more difficult. Fortunately, the two places we do that are the ffma peephole where it doesn't matter and GCM where we already gracefully handle duplicates visits to an instruction. Another aspect here is that using linked lists in this way can be tricky to get right. With sets, things were quite forgiving and the worst that happened if you didn't properly remove a use was that it would get caught in the validator. With linked lists, it can lead to linked list corruption which can be harder to track. However, we do just as much validation of the linked lists as we did of the sets so the validator should still catch these problems. While working on this series, the vast majority of the bugs I had to fix were caught by assertions. I don't think the lists are going to be that much worse than the sets. Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
2015-04-24 10:16:27 -07:00
_mesa_set_remove(reg_state->defs, entry);
}
if (reg_state->defs->entries != 0) {
printf("extra entries in register defs:\n");
struct set_entry *entry;
nir/nir: Use a linked list instead of a hash set for use/def sets This commit switches us from the current setup of using hash sets for use/def sets to using linked lists. Doing so should save us quite a bit of memory because we aren't carrying around 3 hash sets per register and 2 per SSA value. It should also save us CPU time because adding/removing things from use/def sets is 4 pointer manipulations instead of a hash lookup. Running shader-db 50 times with USE_NIR=0, NIR, and NIR + use/def lists: GLSL IR Only: 586.4 +/- 1.653833 NIR with hash sets: 675.4 +/- 2.502108 NIR + use/def lists: 641.2 +/- 1.557043 I also ran a memory usage experiment with Ken's patch to delete GLSL IR and keep NIR. This patch cuts an aditional 42.9 MiB of ralloc'd memory over and above what we gained by deleting the GLSL IR on the same dota trace. On the code complexity side of things, some things are now much easier and others are a bit harder. One of the operations we perform constantly in optimization passes is to replace one source with another. Due to the fact that an instruction can use the same SSA value multiple times, we had to iterate through the sources of the instruction and determine if the use we were replacing was the only one before removing it from the set of uses. With this patch, uses are per-source not per-instruction so we can just remove it safely. On the other hand, trying to iterate over all of the instructions that use a given value is more difficult. Fortunately, the two places we do that are the ffma peephole where it doesn't matter and GCM where we already gracefully handle duplicates visits to an instruction. Another aspect here is that using linked lists in this way can be tricky to get right. With sets, things were quite forgiving and the worst that happened if you didn't properly remove a use was that it would get caught in the validator. With linked lists, it can lead to linked list corruption which can be harder to track. However, we do just as much validation of the linked lists as we did of the sets so the validator should still catch these problems. While working on this series, the vast majority of the bugs I had to fix were caught by assertions. I don't think the lists are going to be that much worse than the sets. Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
2015-04-24 10:16:27 -07:00
set_foreach(reg_state->defs, entry)
printf("%p\n", entry->key);
abort();
}
}
static void
validate_var_decl(nir_variable *var, bool is_global, validate_state *state)
{
state->var = var;
validate_assert(state, is_global == nir_variable_is_global(var));
/* Must have exactly one mode set */
validate_assert(state, util_is_power_of_two_nonzero(var->data.mode));
if (var->data.compact) {
/* The "compact" flag is only valid on arrays of scalars. */
assert(glsl_type_is_array(var->type));
const struct glsl_type *type = glsl_get_array_element(var->type);
if (nir_is_per_vertex_io(var, state->shader->info.stage)) {
assert(glsl_type_is_array(type));
assert(glsl_type_is_scalar(glsl_get_array_element(type)));
} else {
assert(glsl_type_is_scalar(type));
}
}
if (var->num_members > 0) {
const struct glsl_type *without_array = glsl_without_array(var->type);
validate_assert(state, glsl_type_is_struct(without_array));
validate_assert(state, var->num_members == glsl_get_length(without_array));
validate_assert(state, var->members != NULL);
}
/*
* TODO validate some things ir_validate.cpp does (requires more GLSL type
* support)
*/
_mesa_hash_table_insert(state->var_defs, var,
is_global ? NULL : state->impl);
state->var = NULL;
}
static bool
postvalidate_ssa_def(nir_ssa_def *def, void *void_state)
{
validate_state *state = void_state;
struct hash_entry *entry = _mesa_hash_table_search(state->ssa_defs, def);
assume(entry);
ssa_def_validate_state *def_state = (ssa_def_validate_state *)entry->data;
nir_foreach_use(src, def) {
nir/nir: Use a linked list instead of a hash set for use/def sets This commit switches us from the current setup of using hash sets for use/def sets to using linked lists. Doing so should save us quite a bit of memory because we aren't carrying around 3 hash sets per register and 2 per SSA value. It should also save us CPU time because adding/removing things from use/def sets is 4 pointer manipulations instead of a hash lookup. Running shader-db 50 times with USE_NIR=0, NIR, and NIR + use/def lists: GLSL IR Only: 586.4 +/- 1.653833 NIR with hash sets: 675.4 +/- 2.502108 NIR + use/def lists: 641.2 +/- 1.557043 I also ran a memory usage experiment with Ken's patch to delete GLSL IR and keep NIR. This patch cuts an aditional 42.9 MiB of ralloc'd memory over and above what we gained by deleting the GLSL IR on the same dota trace. On the code complexity side of things, some things are now much easier and others are a bit harder. One of the operations we perform constantly in optimization passes is to replace one source with another. Due to the fact that an instruction can use the same SSA value multiple times, we had to iterate through the sources of the instruction and determine if the use we were replacing was the only one before removing it from the set of uses. With this patch, uses are per-source not per-instruction so we can just remove it safely. On the other hand, trying to iterate over all of the instructions that use a given value is more difficult. Fortunately, the two places we do that are the ffma peephole where it doesn't matter and GCM where we already gracefully handle duplicates visits to an instruction. Another aspect here is that using linked lists in this way can be tricky to get right. With sets, things were quite forgiving and the worst that happened if you didn't properly remove a use was that it would get caught in the validator. With linked lists, it can lead to linked list corruption which can be harder to track. However, we do just as much validation of the linked lists as we did of the sets so the validator should still catch these problems. While working on this series, the vast majority of the bugs I had to fix were caught by assertions. I don't think the lists are going to be that much worse than the sets. Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
2015-04-24 10:16:27 -07:00
struct set_entry *entry = _mesa_set_search(def_state->uses, src);
validate_assert(state, entry);
nir/nir: Use a linked list instead of a hash set for use/def sets This commit switches us from the current setup of using hash sets for use/def sets to using linked lists. Doing so should save us quite a bit of memory because we aren't carrying around 3 hash sets per register and 2 per SSA value. It should also save us CPU time because adding/removing things from use/def sets is 4 pointer manipulations instead of a hash lookup. Running shader-db 50 times with USE_NIR=0, NIR, and NIR + use/def lists: GLSL IR Only: 586.4 +/- 1.653833 NIR with hash sets: 675.4 +/- 2.502108 NIR + use/def lists: 641.2 +/- 1.557043 I also ran a memory usage experiment with Ken's patch to delete GLSL IR and keep NIR. This patch cuts an aditional 42.9 MiB of ralloc'd memory over and above what we gained by deleting the GLSL IR on the same dota trace. On the code complexity side of things, some things are now much easier and others are a bit harder. One of the operations we perform constantly in optimization passes is to replace one source with another. Due to the fact that an instruction can use the same SSA value multiple times, we had to iterate through the sources of the instruction and determine if the use we were replacing was the only one before removing it from the set of uses. With this patch, uses are per-source not per-instruction so we can just remove it safely. On the other hand, trying to iterate over all of the instructions that use a given value is more difficult. Fortunately, the two places we do that are the ffma peephole where it doesn't matter and GCM where we already gracefully handle duplicates visits to an instruction. Another aspect here is that using linked lists in this way can be tricky to get right. With sets, things were quite forgiving and the worst that happened if you didn't properly remove a use was that it would get caught in the validator. With linked lists, it can lead to linked list corruption which can be harder to track. However, we do just as much validation of the linked lists as we did of the sets so the validator should still catch these problems. While working on this series, the vast majority of the bugs I had to fix were caught by assertions. I don't think the lists are going to be that much worse than the sets. Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
2015-04-24 10:16:27 -07:00
_mesa_set_remove(def_state->uses, entry);
}
nir/nir: Use a linked list instead of a hash set for use/def sets This commit switches us from the current setup of using hash sets for use/def sets to using linked lists. Doing so should save us quite a bit of memory because we aren't carrying around 3 hash sets per register and 2 per SSA value. It should also save us CPU time because adding/removing things from use/def sets is 4 pointer manipulations instead of a hash lookup. Running shader-db 50 times with USE_NIR=0, NIR, and NIR + use/def lists: GLSL IR Only: 586.4 +/- 1.653833 NIR with hash sets: 675.4 +/- 2.502108 NIR + use/def lists: 641.2 +/- 1.557043 I also ran a memory usage experiment with Ken's patch to delete GLSL IR and keep NIR. This patch cuts an aditional 42.9 MiB of ralloc'd memory over and above what we gained by deleting the GLSL IR on the same dota trace. On the code complexity side of things, some things are now much easier and others are a bit harder. One of the operations we perform constantly in optimization passes is to replace one source with another. Due to the fact that an instruction can use the same SSA value multiple times, we had to iterate through the sources of the instruction and determine if the use we were replacing was the only one before removing it from the set of uses. With this patch, uses are per-source not per-instruction so we can just remove it safely. On the other hand, trying to iterate over all of the instructions that use a given value is more difficult. Fortunately, the two places we do that are the ffma peephole where it doesn't matter and GCM where we already gracefully handle duplicates visits to an instruction. Another aspect here is that using linked lists in this way can be tricky to get right. With sets, things were quite forgiving and the worst that happened if you didn't properly remove a use was that it would get caught in the validator. With linked lists, it can lead to linked list corruption which can be harder to track. However, we do just as much validation of the linked lists as we did of the sets so the validator should still catch these problems. While working on this series, the vast majority of the bugs I had to fix were caught by assertions. I don't think the lists are going to be that much worse than the sets. Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
2015-04-24 10:16:27 -07:00
if (def_state->uses->entries != 0) {
printf("extra entries in SSA def uses:\n");
nir/nir: Use a linked list instead of a hash set for use/def sets This commit switches us from the current setup of using hash sets for use/def sets to using linked lists. Doing so should save us quite a bit of memory because we aren't carrying around 3 hash sets per register and 2 per SSA value. It should also save us CPU time because adding/removing things from use/def sets is 4 pointer manipulations instead of a hash lookup. Running shader-db 50 times with USE_NIR=0, NIR, and NIR + use/def lists: GLSL IR Only: 586.4 +/- 1.653833 NIR with hash sets: 675.4 +/- 2.502108 NIR + use/def lists: 641.2 +/- 1.557043 I also ran a memory usage experiment with Ken's patch to delete GLSL IR and keep NIR. This patch cuts an aditional 42.9 MiB of ralloc'd memory over and above what we gained by deleting the GLSL IR on the same dota trace. On the code complexity side of things, some things are now much easier and others are a bit harder. One of the operations we perform constantly in optimization passes is to replace one source with another. Due to the fact that an instruction can use the same SSA value multiple times, we had to iterate through the sources of the instruction and determine if the use we were replacing was the only one before removing it from the set of uses. With this patch, uses are per-source not per-instruction so we can just remove it safely. On the other hand, trying to iterate over all of the instructions that use a given value is more difficult. Fortunately, the two places we do that are the ffma peephole where it doesn't matter and GCM where we already gracefully handle duplicates visits to an instruction. Another aspect here is that using linked lists in this way can be tricky to get right. With sets, things were quite forgiving and the worst that happened if you didn't properly remove a use was that it would get caught in the validator. With linked lists, it can lead to linked list corruption which can be harder to track. However, we do just as much validation of the linked lists as we did of the sets so the validator should still catch these problems. While working on this series, the vast majority of the bugs I had to fix were caught by assertions. I don't think the lists are going to be that much worse than the sets. Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
2015-04-24 10:16:27 -07:00
struct set_entry *entry;
set_foreach(def_state->uses, entry)
printf("%p\n", entry->key);
abort();
}
nir_foreach_if_use(src, def) {
nir/nir: Use a linked list instead of a hash set for use/def sets This commit switches us from the current setup of using hash sets for use/def sets to using linked lists. Doing so should save us quite a bit of memory because we aren't carrying around 3 hash sets per register and 2 per SSA value. It should also save us CPU time because adding/removing things from use/def sets is 4 pointer manipulations instead of a hash lookup. Running shader-db 50 times with USE_NIR=0, NIR, and NIR + use/def lists: GLSL IR Only: 586.4 +/- 1.653833 NIR with hash sets: 675.4 +/- 2.502108 NIR + use/def lists: 641.2 +/- 1.557043 I also ran a memory usage experiment with Ken's patch to delete GLSL IR and keep NIR. This patch cuts an aditional 42.9 MiB of ralloc'd memory over and above what we gained by deleting the GLSL IR on the same dota trace. On the code complexity side of things, some things are now much easier and others are a bit harder. One of the operations we perform constantly in optimization passes is to replace one source with another. Due to the fact that an instruction can use the same SSA value multiple times, we had to iterate through the sources of the instruction and determine if the use we were replacing was the only one before removing it from the set of uses. With this patch, uses are per-source not per-instruction so we can just remove it safely. On the other hand, trying to iterate over all of the instructions that use a given value is more difficult. Fortunately, the two places we do that are the ffma peephole where it doesn't matter and GCM where we already gracefully handle duplicates visits to an instruction. Another aspect here is that using linked lists in this way can be tricky to get right. With sets, things were quite forgiving and the worst that happened if you didn't properly remove a use was that it would get caught in the validator. With linked lists, it can lead to linked list corruption which can be harder to track. However, we do just as much validation of the linked lists as we did of the sets so the validator should still catch these problems. While working on this series, the vast majority of the bugs I had to fix were caught by assertions. I don't think the lists are going to be that much worse than the sets. Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
2015-04-24 10:16:27 -07:00
struct set_entry *entry = _mesa_set_search(def_state->if_uses, src);
validate_assert(state, entry);
nir/nir: Use a linked list instead of a hash set for use/def sets This commit switches us from the current setup of using hash sets for use/def sets to using linked lists. Doing so should save us quite a bit of memory because we aren't carrying around 3 hash sets per register and 2 per SSA value. It should also save us CPU time because adding/removing things from use/def sets is 4 pointer manipulations instead of a hash lookup. Running shader-db 50 times with USE_NIR=0, NIR, and NIR + use/def lists: GLSL IR Only: 586.4 +/- 1.653833 NIR with hash sets: 675.4 +/- 2.502108 NIR + use/def lists: 641.2 +/- 1.557043 I also ran a memory usage experiment with Ken's patch to delete GLSL IR and keep NIR. This patch cuts an aditional 42.9 MiB of ralloc'd memory over and above what we gained by deleting the GLSL IR on the same dota trace. On the code complexity side of things, some things are now much easier and others are a bit harder. One of the operations we perform constantly in optimization passes is to replace one source with another. Due to the fact that an instruction can use the same SSA value multiple times, we had to iterate through the sources of the instruction and determine if the use we were replacing was the only one before removing it from the set of uses. With this patch, uses are per-source not per-instruction so we can just remove it safely. On the other hand, trying to iterate over all of the instructions that use a given value is more difficult. Fortunately, the two places we do that are the ffma peephole where it doesn't matter and GCM where we already gracefully handle duplicates visits to an instruction. Another aspect here is that using linked lists in this way can be tricky to get right. With sets, things were quite forgiving and the worst that happened if you didn't properly remove a use was that it would get caught in the validator. With linked lists, it can lead to linked list corruption which can be harder to track. However, we do just as much validation of the linked lists as we did of the sets so the validator should still catch these problems. While working on this series, the vast majority of the bugs I had to fix were caught by assertions. I don't think the lists are going to be that much worse than the sets. Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
2015-04-24 10:16:27 -07:00
_mesa_set_remove(def_state->if_uses, entry);
}
nir/nir: Use a linked list instead of a hash set for use/def sets This commit switches us from the current setup of using hash sets for use/def sets to using linked lists. Doing so should save us quite a bit of memory because we aren't carrying around 3 hash sets per register and 2 per SSA value. It should also save us CPU time because adding/removing things from use/def sets is 4 pointer manipulations instead of a hash lookup. Running shader-db 50 times with USE_NIR=0, NIR, and NIR + use/def lists: GLSL IR Only: 586.4 +/- 1.653833 NIR with hash sets: 675.4 +/- 2.502108 NIR + use/def lists: 641.2 +/- 1.557043 I also ran a memory usage experiment with Ken's patch to delete GLSL IR and keep NIR. This patch cuts an aditional 42.9 MiB of ralloc'd memory over and above what we gained by deleting the GLSL IR on the same dota trace. On the code complexity side of things, some things are now much easier and others are a bit harder. One of the operations we perform constantly in optimization passes is to replace one source with another. Due to the fact that an instruction can use the same SSA value multiple times, we had to iterate through the sources of the instruction and determine if the use we were replacing was the only one before removing it from the set of uses. With this patch, uses are per-source not per-instruction so we can just remove it safely. On the other hand, trying to iterate over all of the instructions that use a given value is more difficult. Fortunately, the two places we do that are the ffma peephole where it doesn't matter and GCM where we already gracefully handle duplicates visits to an instruction. Another aspect here is that using linked lists in this way can be tricky to get right. With sets, things were quite forgiving and the worst that happened if you didn't properly remove a use was that it would get caught in the validator. With linked lists, it can lead to linked list corruption which can be harder to track. However, we do just as much validation of the linked lists as we did of the sets so the validator should still catch these problems. While working on this series, the vast majority of the bugs I had to fix were caught by assertions. I don't think the lists are going to be that much worse than the sets. Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
2015-04-24 10:16:27 -07:00
if (def_state->if_uses->entries != 0) {
printf("extra entries in SSA def uses:\n");
nir/nir: Use a linked list instead of a hash set for use/def sets This commit switches us from the current setup of using hash sets for use/def sets to using linked lists. Doing so should save us quite a bit of memory because we aren't carrying around 3 hash sets per register and 2 per SSA value. It should also save us CPU time because adding/removing things from use/def sets is 4 pointer manipulations instead of a hash lookup. Running shader-db 50 times with USE_NIR=0, NIR, and NIR + use/def lists: GLSL IR Only: 586.4 +/- 1.653833 NIR with hash sets: 675.4 +/- 2.502108 NIR + use/def lists: 641.2 +/- 1.557043 I also ran a memory usage experiment with Ken's patch to delete GLSL IR and keep NIR. This patch cuts an aditional 42.9 MiB of ralloc'd memory over and above what we gained by deleting the GLSL IR on the same dota trace. On the code complexity side of things, some things are now much easier and others are a bit harder. One of the operations we perform constantly in optimization passes is to replace one source with another. Due to the fact that an instruction can use the same SSA value multiple times, we had to iterate through the sources of the instruction and determine if the use we were replacing was the only one before removing it from the set of uses. With this patch, uses are per-source not per-instruction so we can just remove it safely. On the other hand, trying to iterate over all of the instructions that use a given value is more difficult. Fortunately, the two places we do that are the ffma peephole where it doesn't matter and GCM where we already gracefully handle duplicates visits to an instruction. Another aspect here is that using linked lists in this way can be tricky to get right. With sets, things were quite forgiving and the worst that happened if you didn't properly remove a use was that it would get caught in the validator. With linked lists, it can lead to linked list corruption which can be harder to track. However, we do just as much validation of the linked lists as we did of the sets so the validator should still catch these problems. While working on this series, the vast majority of the bugs I had to fix were caught by assertions. I don't think the lists are going to be that much worse than the sets. Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
2015-04-24 10:16:27 -07:00
struct set_entry *entry;
set_foreach(def_state->if_uses, entry)
printf("%p\n", entry->key);
abort();
}
return true;
}
static void
validate_function_impl(nir_function_impl *impl, validate_state *state)
{
validate_assert(state, impl->function->impl == impl);
validate_assert(state, impl->cf_node.parent == NULL);
validate_assert(state, exec_list_is_empty(&impl->end_block->instr_list));
validate_assert(state, impl->end_block->successors[0] == NULL);
validate_assert(state, impl->end_block->successors[1] == NULL);
state->impl = impl;
state->parent_node = &impl->cf_node;
exec_list_validate(&impl->locals);
nir_foreach_variable(var, &impl->locals) {
validate_var_decl(var, false, state);
}
state->regs_found = realloc(state->regs_found,
BITSET_WORDS(impl->reg_alloc) *
sizeof(BITSET_WORD));
memset(state->regs_found, 0, BITSET_WORDS(impl->reg_alloc) *
sizeof(BITSET_WORD));
exec_list_validate(&impl->registers);
foreach_list_typed(nir_register, reg, node, &impl->registers) {
prevalidate_reg_decl(reg, false, state);
}
state->ssa_defs_found = realloc(state->ssa_defs_found,
BITSET_WORDS(impl->ssa_alloc) *
sizeof(BITSET_WORD));
memset(state->ssa_defs_found, 0, BITSET_WORDS(impl->ssa_alloc) *
sizeof(BITSET_WORD));
exec_list_validate(&impl->body);
foreach_list_typed(nir_cf_node, node, node, &impl->body) {
validate_cf_node(node, state);
}
foreach_list_typed(nir_register, reg, node, &impl->registers) {
postvalidate_reg_decl(reg, state);
}
nir_foreach_block(block, impl) {
nir_foreach_instr(instr, block)
nir_foreach_ssa_def(instr, postvalidate_ssa_def, state);
}
}
static void
validate_function(nir_function *func, validate_state *state)
{
if (func->impl != NULL) {
validate_assert(state, func->impl->function == func);
validate_function_impl(func->impl, state);
}
}
static void
init_validate_state(validate_state *state)
{
state->regs = _mesa_hash_table_create(NULL, _mesa_hash_pointer,
_mesa_key_pointer_equal);
state->ssa_defs = _mesa_hash_table_create(NULL, _mesa_hash_pointer,
_mesa_key_pointer_equal);
state->ssa_defs_found = NULL;
state->regs_found = NULL;
state->var_defs = _mesa_hash_table_create(NULL, _mesa_hash_pointer,
_mesa_key_pointer_equal);
state->errors = _mesa_hash_table_create(NULL, _mesa_hash_pointer,
_mesa_key_pointer_equal);
state->loop = NULL;
state->instr = NULL;
state->var = NULL;
}
static void
destroy_validate_state(validate_state *state)
{
_mesa_hash_table_destroy(state->regs, NULL);
_mesa_hash_table_destroy(state->ssa_defs, NULL);
free(state->ssa_defs_found);
free(state->regs_found);
_mesa_hash_table_destroy(state->var_defs, NULL);
_mesa_hash_table_destroy(state->errors, NULL);
}
static void
dump_errors(validate_state *state)
{
struct hash_table *errors = state->errors;
fprintf(stderr, "%d errors:\n", _mesa_hash_table_num_entries(errors));
nir_print_shader_annotated(state->shader, stderr, errors);
if (_mesa_hash_table_num_entries(errors) > 0) {
fprintf(stderr, "%d additional errors:\n",
_mesa_hash_table_num_entries(errors));
struct hash_entry *entry;
hash_table_foreach(errors, entry) {
fprintf(stderr, "%s\n", (char *)entry->data);
}
}
abort();
}
void
nir_validate_shader(nir_shader *shader)
{
static int should_validate = -1;
if (should_validate < 0)
should_validate = env_var_as_boolean("NIR_VALIDATE", true);
if (!should_validate)
return;
validate_state state;
init_validate_state(&state);
state.shader = shader;
exec_list_validate(&shader->uniforms);
nir_foreach_variable(var, &shader->uniforms) {
validate_var_decl(var, true, &state);
}
exec_list_validate(&shader->inputs);
nir_foreach_variable(var, &shader->inputs) {
validate_var_decl(var, true, &state);
}
exec_list_validate(&shader->outputs);
nir_foreach_variable(var, &shader->outputs) {
validate_var_decl(var, true, &state);
}
exec_list_validate(&shader->shared);
nir_foreach_variable(var, &shader->shared) {
validate_var_decl(var, true, &state);
}
exec_list_validate(&shader->globals);
nir_foreach_variable(var, &shader->globals) {
validate_var_decl(var, true, &state);
}
exec_list_validate(&shader->system_values);
nir_foreach_variable(var, &shader->system_values) {
validate_var_decl(var, true, &state);
}
state.regs_found = realloc(state.regs_found,
BITSET_WORDS(shader->reg_alloc) *
sizeof(BITSET_WORD));
memset(state.regs_found, 0, BITSET_WORDS(shader->reg_alloc) *
sizeof(BITSET_WORD));
exec_list_validate(&shader->registers);
foreach_list_typed(nir_register, reg, node, &shader->registers) {
prevalidate_reg_decl(reg, true, &state);
}
exec_list_validate(&shader->functions);
foreach_list_typed(nir_function, func, node, &shader->functions) {
validate_function(func, &state);
}
foreach_list_typed(nir_register, reg, node, &shader->registers) {
postvalidate_reg_decl(reg, &state);
}
if (_mesa_hash_table_num_entries(state.errors) > 0)
dump_errors(&state);
destroy_validate_state(&state);
}
#endif /* NDEBUG */