summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
Diffstat
-rw-r--r--libcuda/gpgpu_context.h2
-rw-r--r--src/abstract_hardware_model.cc13
-rw-r--r--src/abstract_hardware_model.h22
-rw-r--r--src/gpgpu-sim/dram.cc2
-rw-r--r--src/gpgpu-sim/dram.h5
-rw-r--r--src/gpgpu-sim/gpu-cache.cc6
-rw-r--r--src/gpgpu-sim/gpu-sim.cc2
-rw-r--r--src/gpgpu-sim/gpu-sim.h15
-rw-r--r--src/gpgpu-sim/l2cache.cc11
-rw-r--r--src/gpgpu-sim/l2cache.h12
-rw-r--r--src/gpgpu-sim/mem_fetch.cc4
-rw-r--r--src/gpgpu-sim/mem_fetch.h5
-rw-r--r--src/gpgpu-sim/mem_latency_stat.cc2
-rw-r--r--src/gpgpu-sim/mem_latency_stat.h5
-rw-r--r--src/gpgpu-sim/power_stat.cc4
-rw-r--r--src/gpgpu-sim/power_stat.h6
-rw-r--r--src/gpgpu-sim/shader.cc20
-rw-r--r--src/gpgpu-sim/shader.h20
18 files changed, 85 insertions, 71 deletions
diff --git a/libcuda/gpgpu_context.h b/libcuda/gpgpu_context.h
index c0b250a..07473be 100644
--- a/libcuda/gpgpu_context.h
+++ b/libcuda/gpgpu_context.h
@@ -11,6 +11,7 @@ class gpgpu_context {
public:
gpgpu_context() {
g_global_allfiles_symbol_table = NULL;
+ sm_next_access_uid=0;
api = new cuda_runtime_api(this);
ptxinfo = new ptxinfo_data(this);
ptx_parser = new ptx_recognizer(this);
@@ -21,6 +22,7 @@ class gpgpu_context {
// global list
symbol_table *g_global_allfiles_symbol_table;
const char *g_filename;
+ unsigned sm_next_access_uid;
// objects pointers for each file
cuda_runtime_api* api;
ptxinfo_data* ptxinfo;
diff --git a/src/abstract_hardware_model.cc b/src/abstract_hardware_model.cc
index fde7874..fe10daa 100644
--- a/src/abstract_hardware_model.cc
+++ b/src/abstract_hardware_model.cc
@@ -41,9 +41,16 @@
#include <iostream>
#include "../libcuda/gpgpu_context.h"
-unsigned mem_access_t::sm_next_access_uid = 0;
unsigned warp_inst_t::sm_next_uid = 0;
+void mem_access_t::init(gpgpu_context* ctx)
+{
+ gpgpu_ctx = ctx;
+ m_uid=++(gpgpu_ctx->sm_next_access_uid);
+ m_addr=0;
+ m_req_size=0;
+}
+
checkpoint::checkpoint()
{
@@ -449,7 +456,7 @@ void warp_inst_t::generate_mem_accesses()
byte_mask.set(idx+i);
}
for( a=accesses.begin(); a != accesses.end(); ++a )
- m_accessq.push_back( mem_access_t(access_type,a->first,cache_block_size,is_write,a->second, byte_mask, mem_access_sector_mask_t()));
+ m_accessq.push_back( mem_access_t(access_type,a->first,cache_block_size,is_write,a->second, byte_mask, mem_access_sector_mask_t(), m_config->gpgpu_ctx));
}
if ( space.get_type() == global_space ) {
@@ -681,7 +688,7 @@ void warp_inst_t::memory_coalescing_arch_reduce_and_send( bool is_write, mem_acc
assert(lower_half_used && upper_half_used);
}
}
- m_accessq.push_back( mem_access_t(access_type,addr,size,is_write,info.active,info.bytes, info.chunks) );
+ m_accessq.push_back( mem_access_t(access_type,addr,size,is_write,info.active,info.bytes, info.chunks,m_config->gpgpu_ctx) );
}
void warp_inst_t::completed( unsigned long long cycle ) const
diff --git a/src/abstract_hardware_model.h b/src/abstract_hardware_model.h
index 8ef8376..a22c8c3 100644
--- a/src/abstract_hardware_model.h
+++ b/src/abstract_hardware_model.h
@@ -732,13 +732,14 @@ enum cache_operator_type {
class mem_access_t {
public:
- mem_access_t() { init(); }
+ mem_access_t(gpgpu_context* ctx) { init(ctx); }
mem_access_t( mem_access_type type,
new_addr_type address,
unsigned size,
- bool wr )
+ bool wr,
+ gpgpu_context* ctx)
{
- init();
+ init(ctx);
m_type = type;
m_addr = address;
m_req_size = size;
@@ -750,10 +751,11 @@ public:
bool wr,
const active_mask_t &active_mask,
const mem_access_byte_mask_t &byte_mask,
- const mem_access_sector_mask_t &sector_mask)
+ const mem_access_sector_mask_t &sector_mask,
+ gpgpu_context* ctx)
: m_warp_mask(active_mask), m_byte_mask(byte_mask), m_sector_mask(sector_mask)
{
- init();
+ init(ctx);
m_type = type;
m_addr = address;
m_req_size = size;
@@ -786,13 +788,9 @@ public:
}
}
+ gpgpu_context* gpgpu_ctx;
private:
- void init()
- {
- m_uid=++sm_next_access_uid;
- m_addr=0;
- m_req_size=0;
- }
+ void init(gpgpu_context* ctx);
unsigned m_uid;
new_addr_type m_addr; // request address
@@ -802,8 +800,6 @@ private:
active_mask_t m_warp_mask;
mem_access_byte_mask_t m_byte_mask;
mem_access_sector_mask_t m_sector_mask;
-
- static unsigned sm_next_access_uid;
};
class mem_fetch;
diff --git a/src/gpgpu-sim/dram.cc b/src/gpgpu-sim/dram.cc
index 5e36d4b..d443d79 100644
--- a/src/gpgpu-sim/dram.cc
+++ b/src/gpgpu-sim/dram.cc
@@ -41,7 +41,7 @@ int PRINT_CYCLE = 0;
template class fifo_pipeline<mem_fetch>;
template class fifo_pipeline<dram_req_t>;
-dram_t::dram_t( unsigned int partition_id, const struct memory_config *config, memory_stats_t *stats,
+dram_t::dram_t( unsigned int partition_id, const memory_config *config, memory_stats_t *stats,
memory_partition_unit *mp, gpgpu_sim* gpu )
{
id = partition_id;
diff --git a/src/gpgpu-sim/dram.h b/src/gpgpu-sim/dram.h
index 7a3a2da..0bd9725 100644
--- a/src/gpgpu-sim/dram.h
+++ b/src/gpgpu-sim/dram.h
@@ -106,11 +106,12 @@ enum bank_grp_bits_position{
};
class mem_fetch;
+class memory_config;
class dram_t
{
public:
- dram_t( unsigned int parition_id, const struct memory_config *config, class memory_stats_t *stats,
+ dram_t( unsigned int parition_id, const memory_config *config, class memory_stats_t *stats,
class memory_partition_unit *mp, class gpgpu_sim* gpu );
bool full(bool is_write) const;
@@ -145,7 +146,7 @@ public:
- const struct memory_config *m_config;
+ const memory_config *m_config;
private:
bankgrp_t **bkgrp;
diff --git a/src/gpgpu-sim/gpu-cache.cc b/src/gpgpu-sim/gpu-cache.cc
index f1f6e19..1705821 100644
--- a/src/gpgpu-sim/gpu-cache.cc
+++ b/src/gpgpu-sim/gpu-cache.cc
@@ -1251,7 +1251,8 @@ data_cache::wr_miss_wa_naive( new_addr_type addr,
false, // Now performing a read
mf->get_access_warp_mask(),
mf->get_access_byte_mask(),
- mf->get_access_sector_mask());
+ mf->get_access_sector_mask(),
+ m_gpu->gpgpu_ctx);
mem_fetch *n_mf = new mem_fetch( *ma,
NULL,
@@ -1365,7 +1366,8 @@ data_cache::wr_miss_wa_fetch_on_write( new_addr_type addr,
false, // Now performing a read
mf->get_access_warp_mask(),
mf->get_access_byte_mask(),
- mf->get_access_sector_mask());
+ mf->get_access_sector_mask(),
+ m_gpu->gpgpu_ctx);
mem_fetch *n_mf = new mem_fetch( *ma,
NULL,
diff --git a/src/gpgpu-sim/gpu-sim.cc b/src/gpgpu-sim/gpu-sim.cc
index 0481259..0644b44 100644
--- a/src/gpgpu-sim/gpu-sim.cc
+++ b/src/gpgpu-sim/gpu-sim.cc
@@ -1828,7 +1828,7 @@ const shader_core_config * gpgpu_sim::getShaderCoreConfig()
return m_shader_config;
}
-const struct memory_config * gpgpu_sim::getMemoryConfig()
+const memory_config * gpgpu_sim::getMemoryConfig()
{
return m_memory_config;
}
diff --git a/src/gpgpu-sim/gpu-sim.h b/src/gpgpu-sim/gpu-sim.h
index 78f0505..19e1eb3 100644
--- a/src/gpgpu-sim/gpu-sim.h
+++ b/src/gpgpu-sim/gpu-sim.h
@@ -33,6 +33,7 @@
#include "../trace.h"
#include "addrdec.h"
#include "shader.h"
+#include "gpu-cache.h"
#include <iostream>
#include <fstream>
#include <list>
@@ -143,13 +144,14 @@ struct power_config {
};
-
-struct memory_config {
- memory_config()
+class memory_config {
+ public:
+ memory_config(gpgpu_context* ctx)
{
m_valid = false;
gpgpu_dram_timing_opt=NULL;
gpgpu_L2_queue_config=NULL;
+ gpgpu_ctx = ctx;
}
void init()
{
@@ -291,13 +293,14 @@ struct memory_config {
unsigned write_high_watermark;
unsigned write_low_watermark;
bool m_perf_sim_memcpy;
+ gpgpu_context* gpgpu_ctx;
};
extern bool g_interactive_debugger_enabled;
class gpgpu_sim_config : public power_config, public gpgpu_functional_sim_config {
public:
- gpgpu_sim_config(gpgpu_context* ctx): m_shader_config(ctx) {
+ gpgpu_sim_config(gpgpu_context* ctx): m_shader_config(ctx), m_memory_config(ctx) {
m_valid = false;
gpgpu_ctx = ctx;
}
@@ -507,7 +510,7 @@ public:
/*!
* Returning the memory configuration of the shader core, used by the functional simulation only so far
*/
- const struct memory_config * getMemoryConfig();
+ const memory_config * getMemoryConfig();
//! Get shader core SIMT cluster
@@ -567,7 +570,7 @@ private:
const struct cudaDeviceProp *m_cuda_properties;
const shader_core_config *m_shader_config;
- const struct memory_config *m_memory_config;
+ const memory_config *m_memory_config;
// stats
class shader_core_stats *m_shader_stats;
diff --git a/src/gpgpu-sim/l2cache.cc b/src/gpgpu-sim/l2cache.cc
index 62e70a7..6540b52 100644
--- a/src/gpgpu-sim/l2cache.cc
+++ b/src/gpgpu-sim/l2cache.cc
@@ -49,7 +49,7 @@
mem_fetch * partition_mf_allocator::alloc(new_addr_type addr, mem_access_type type, unsigned size, bool wr, unsigned long long cycle ) const
{
assert( wr );
- mem_access_t access( type, addr, size, wr );
+ mem_access_t access( type, addr, size, wr, m_memory_config->gpgpu_ctx );
mem_fetch *mf = new mem_fetch( access,
NULL,
WRITE_PACKET_SIZE,
@@ -62,7 +62,7 @@ mem_fetch * partition_mf_allocator::alloc(new_addr_type addr, mem_access_type ty
}
memory_partition_unit::memory_partition_unit( unsigned partition_id,
- const struct memory_config *config,
+ const memory_config *config,
class memory_stats_t *stats,
class gpgpu_sim* gpu)
: m_id(partition_id), m_config(config), m_stats(stats), m_arbitration_metadata(config), m_gpu(gpu)
@@ -95,7 +95,7 @@ memory_partition_unit::~memory_partition_unit()
delete[] m_sub_partition;
}
-memory_partition_unit::arbitration_metadata::arbitration_metadata(const struct memory_config *config)
+memory_partition_unit::arbitration_metadata::arbitration_metadata(const memory_config *config)
: m_last_borrower(config->m_n_sub_partition_per_memory_channel - 1),
m_private_credit(config->m_n_sub_partition_per_memory_channel, 0),
m_shared_credit(0)
@@ -312,7 +312,7 @@ void memory_partition_unit::print( FILE *fp ) const
}
memory_sub_partition::memory_sub_partition( unsigned sub_partition_id,
- const struct memory_config *config,
+ const memory_config *config,
class memory_stats_t *stats,
class gpgpu_sim* gpu)
{
@@ -640,7 +640,8 @@ std::vector<mem_fetch*> memory_sub_partition::breakdown_request_to_sector_reques
mf->is_write(),
mf->get_access_warp_mask(),
mf->get_access_byte_mask() & byte_sector_mask,
- std::bitset<SECTOR_CHUNCK_SIZE>().set(j));
+ std::bitset<SECTOR_CHUNCK_SIZE>().set(j),
+ m_gpu->gpgpu_ctx);
mem_fetch *n_mf = new mem_fetch( *ma,
NULL,
diff --git a/src/gpgpu-sim/l2cache.h b/src/gpgpu-sim/l2cache.h
index 8ff2666..1f74c47 100644
--- a/src/gpgpu-sim/l2cache.h
+++ b/src/gpgpu-sim/l2cache.h
@@ -58,7 +58,7 @@ private:
class memory_partition_unit
{
public:
- memory_partition_unit( unsigned partition_id, const struct memory_config *config, class memory_stats_t *stats, class gpgpu_sim* gpu );
+ memory_partition_unit( unsigned partition_id, const memory_config *config, class memory_stats_t *stats, class gpgpu_sim* gpu );
~memory_partition_unit();
bool busy() const;
@@ -98,7 +98,7 @@ public:
private:
unsigned m_id;
- const struct memory_config *m_config;
+ const memory_config *m_config;
class memory_stats_t *m_stats;
class memory_sub_partition **m_sub_partition;
class dram_t *m_dram;
@@ -106,7 +106,7 @@ private:
class arbitration_metadata
{
public:
- arbitration_metadata(const struct memory_config *config);
+ arbitration_metadata(const memory_config *config);
// check if a subpartition still has credit
bool has_credits(int inner_sub_partition_id) const;
@@ -130,7 +130,7 @@ private:
std::vector<int> m_private_credit;
int m_shared_credit;
};
- arbitration_metadata m_arbitration_metadata;
+ arbitration_metadata m_arbitration_metadata;
// determine wheither a given subpartition can issue to DRAM
bool can_issue_to_dram(int inner_sub_partition_id);
@@ -149,7 +149,7 @@ private:
class memory_sub_partition
{
public:
- memory_sub_partition( unsigned sub_partition_id, const struct memory_config *config, class memory_stats_t *stats, class gpgpu_sim* gpu );
+ memory_sub_partition( unsigned sub_partition_id, const memory_config *config, class memory_stats_t *stats, class gpgpu_sim* gpu );
~memory_sub_partition();
unsigned get_id() const { return m_id; }
@@ -197,7 +197,7 @@ public:
private:
// data
unsigned m_id; //< the global sub partition ID
- const struct memory_config *m_config;
+ const memory_config *m_config;
class l2_cache *m_L2cache;
class L2interface *m_L2interface;
class gpgpu_sim* m_gpu;
diff --git a/src/gpgpu-sim/mem_fetch.cc b/src/gpgpu-sim/mem_fetch.cc
index c9b0484..6a00889 100644
--- a/src/gpgpu-sim/mem_fetch.cc
+++ b/src/gpgpu-sim/mem_fetch.cc
@@ -39,10 +39,10 @@ mem_fetch::mem_fetch( const mem_access_t &access,
unsigned wid,
unsigned sid,
unsigned tpc,
- const struct memory_config *config,
+ const memory_config *config,
unsigned long long cycle,
mem_fetch *m_original_mf,
- mem_fetch *m_original_wr_mf)
+ mem_fetch *m_original_wr_mf):m_access(access)
{
m_request_uid = sm_next_mf_request_uid++;
diff --git a/src/gpgpu-sim/mem_fetch.h b/src/gpgpu-sim/mem_fetch.h
index 4eb3a52..1cab9f2 100644
--- a/src/gpgpu-sim/mem_fetch.h
+++ b/src/gpgpu-sim/mem_fetch.h
@@ -47,6 +47,7 @@ enum mf_type {
#undef MF_TUP
#undef MF_TUP_END
+class memory_config;
class mem_fetch {
public:
mem_fetch( const mem_access_t &access,
@@ -55,7 +56,7 @@ public:
unsigned wid,
unsigned sid,
unsigned tpc,
- const struct memory_config *config,
+ const memory_config *config,
unsigned long long cycle,
mem_fetch *original_mf = NULL,
mem_fetch *original_wr_mf = NULL);
@@ -149,7 +150,7 @@ private:
static unsigned sm_next_mf_request_uid;
- const struct memory_config *m_mem_config;
+ const memory_config *m_mem_config;
unsigned icnt_flit_size;
mem_fetch* original_mf; //this pointer is set up when a request is divided into sector requests at L2 cache (if the req size > L2 sector size), so the pointer refers to the original request
diff --git a/src/gpgpu-sim/mem_latency_stat.cc b/src/gpgpu-sim/mem_latency_stat.cc
index d08ba39..4e94991 100644
--- a/src/gpgpu-sim/mem_latency_stat.cc
+++ b/src/gpgpu-sim/mem_latency_stat.cc
@@ -42,7 +42,7 @@
#include <stdlib.h>
#include <stdio.h>
-memory_stats_t::memory_stats_t( unsigned n_shader, const shader_core_config *shader_config, const struct memory_config *mem_config, const class gpgpu_sim* gpu )
+memory_stats_t::memory_stats_t( unsigned n_shader, const shader_core_config *shader_config, const memory_config *mem_config, const class gpgpu_sim* gpu )
{
assert( mem_config->m_valid );
assert( shader_config->m_valid );
diff --git a/src/gpgpu-sim/mem_latency_stat.h b/src/gpgpu-sim/mem_latency_stat.h
index 6ce568d..0c84972 100644
--- a/src/gpgpu-sim/mem_latency_stat.h
+++ b/src/gpgpu-sim/mem_latency_stat.h
@@ -32,11 +32,12 @@
#include <zlib.h>
#include <map>
+class memory_config;
class memory_stats_t {
public:
memory_stats_t( unsigned n_shader,
const class shader_core_config *shader_config,
- const struct memory_config *mem_config,
+ const memory_config *mem_config,
const class gpgpu_sim* gpu);
unsigned memlatstat_done( class mem_fetch *mf );
@@ -54,7 +55,7 @@ public:
unsigned m_n_shader;
const shader_core_config *m_shader_config;
- const struct memory_config *m_memory_config;
+ const memory_config *m_memory_config;
const class gpgpu_sim* m_gpu;
unsigned max_mrq_latency;
diff --git a/src/gpgpu-sim/power_stat.cc b/src/gpgpu-sim/power_stat.cc
index 007b4c6..2c02082 100644
--- a/src/gpgpu-sim/power_stat.cc
+++ b/src/gpgpu-sim/power_stat.cc
@@ -42,7 +42,7 @@
-power_mem_stat_t::power_mem_stat_t(const struct memory_config *mem_config, const shader_core_config *shdr_config, memory_stats_t *mem_stats, shader_core_stats *shdr_stats){
+power_mem_stat_t::power_mem_stat_t(const memory_config *mem_config, const shader_core_config *shdr_config, memory_stats_t *mem_stats, shader_core_stats *shdr_stats){
assert( mem_config->m_valid );
m_mem_stats = mem_stats;
m_config = mem_config;
@@ -266,7 +266,7 @@ for(unsigned i=0; i<m_config->num_shader(); ++i){
}
}
-power_stat_t::power_stat_t( const shader_core_config *shader_config,float * average_pipeline_duty_cycle,float *active_sms,shader_core_stats * shader_stats, const struct memory_config *mem_config,memory_stats_t * memory_stats)
+power_stat_t::power_stat_t( const shader_core_config *shader_config,float * average_pipeline_duty_cycle,float *active_sms,shader_core_stats * shader_stats, const memory_config *mem_config,memory_stats_t * memory_stats)
{
assert( shader_config->m_valid );
assert( mem_config->m_valid );
diff --git a/src/gpgpu-sim/power_stat.h b/src/gpgpu-sim/power_stat.h
index 91fade9..24ade99 100644
--- a/src/gpgpu-sim/power_stat.h
+++ b/src/gpgpu-sim/power_stat.h
@@ -113,7 +113,7 @@ struct mem_power_stats_pod{
class power_mem_stat_t : public mem_power_stats_pod{
public:
- power_mem_stat_t(const struct memory_config *mem_config, const shader_core_config *shdr_config, memory_stats_t *mem_stats, shader_core_stats *shdr_stats);
+ power_mem_stat_t(const memory_config *mem_config, const shader_core_config *shdr_config, memory_stats_t *mem_stats, shader_core_stats *shdr_stats);
void visualizer_print( gzFile visualizer_file );
void print (FILE *fout) const;
void init();
@@ -128,7 +128,7 @@ private:
class power_stat_t {
public:
- power_stat_t( const shader_core_config *shader_config,float * average_pipeline_duty_cycle,float * active_sms,shader_core_stats * shader_stats, const struct memory_config *mem_config,memory_stats_t * memory_stats);
+ power_stat_t( const shader_core_config *shader_config,float * average_pipeline_duty_cycle,float * active_sms,shader_core_stats * shader_stats, const memory_config *mem_config,memory_stats_t * memory_stats);
void visualizer_print( gzFile visualizer_file );
void print (FILE *fout) const;
void save_stats(){
@@ -621,7 +621,7 @@ public:
float * m_average_pipeline_duty_cycle;
float * m_active_sms;
const shader_core_config *m_config;
- const struct memory_config *m_mem_config;
+ const memory_config *m_mem_config;
};
diff --git a/src/gpgpu-sim/shader.cc b/src/gpgpu-sim/shader.cc
index f380560..b7ae95d 100644
--- a/src/gpgpu-sim/shader.cc
+++ b/src/gpgpu-sim/shader.cc
@@ -28,7 +28,6 @@
#include <float.h>
#include "shader.h"
-#include "gpu-sim.h"
#include "addrdec.h"
#include "dram.h"
#include "stat-tool.h"
@@ -53,6 +52,19 @@
#define MIN(a,b) (((a)<(b))?(a):(b))
+mem_fetch *shader_core_mem_fetch_allocator::alloc( new_addr_type addr, mem_access_type type, unsigned size, bool wr, unsigned long long cycle ) const
+{
+ mem_access_t access( type, addr, size, wr, m_memory_config->gpgpu_ctx);
+ mem_fetch *mf = new mem_fetch( access,
+ NULL,
+ wr?WRITE_PACKET_SIZE:READ_PACKET_SIZE,
+ -1,
+ m_core_id,
+ m_cluster_id,
+ m_memory_config,
+ cycle);
+ return mf;
+}
/////////////////////////////////////////////////////////////////////////////
std::list<unsigned> shader_core_ctx::get_regs_written( const inst_t &fvt ) const
@@ -71,7 +83,7 @@ shader_core_ctx::shader_core_ctx( class gpgpu_sim *gpu,
unsigned shader_id,
unsigned tpc_id,
const shader_core_config *config,
- const struct memory_config *mem_config,
+ const memory_config *mem_config,
shader_core_stats *stats )
: core_t( gpu, NULL, config->warp_size, config->n_thread_per_shader ),
m_barriers( this, config->max_warps_per_shader, config->max_cta_per_core, config->max_barriers_per_cta, config->warp_size ),
@@ -809,7 +821,7 @@ void shader_core_ctx::fetch()
// TODO: replace with use of allocator
// mem_fetch *mf = m_mem_fetch_allocator->alloc()
- mem_access_t acc(INST_ACC_R,ppc,nbytes,false);
+ mem_access_t acc(INST_ACC_R,ppc,nbytes,false, m_gpu->gpgpu_ctx);
mem_fetch *mf = new mem_fetch(acc,
NULL/*we don't have an instruction yet*/,
READ_PACKET_SIZE,
@@ -3787,7 +3799,7 @@ void opndcoll_rfu_t::collector_unit_t::dispatch()
simt_core_cluster::simt_core_cluster( class gpgpu_sim *gpu,
unsigned cluster_id,
const shader_core_config *config,
- const struct memory_config *mem_config,
+ const memory_config *mem_config,
shader_core_stats *stats,
class memory_stats_t *mstats )
{
diff --git a/src/gpgpu-sim/shader.h b/src/gpgpu-sim/shader.h
index 2837f1b..b0d7f7f 100644
--- a/src/gpgpu-sim/shader.h
+++ b/src/gpgpu-sim/shader.h
@@ -1727,6 +1727,7 @@ private:
friend class LooseRoundRobbinScheduler;
};
+class memory_config;
class shader_core_mem_fetch_allocator : public mem_fetch_allocator {
public:
shader_core_mem_fetch_allocator( unsigned core_id, unsigned cluster_id, const memory_config *config )
@@ -1735,20 +1736,7 @@ public:
m_cluster_id = cluster_id;
m_memory_config = config;
}
- mem_fetch *alloc( new_addr_type addr, mem_access_type type, unsigned size, bool wr, unsigned long long cycle ) const
- {
- mem_access_t access( type, addr, size, wr );
- mem_fetch *mf = new mem_fetch( access,
- NULL,
- wr?WRITE_PACKET_SIZE:READ_PACKET_SIZE,
- -1,
- m_core_id,
- m_cluster_id,
- m_memory_config,
- cycle);
- return mf;
- }
-
+ mem_fetch *alloc( new_addr_type addr, mem_access_type type, unsigned size, bool wr, unsigned long long cycle ) const;
mem_fetch *alloc( const warp_inst_t &inst, const mem_access_t &access, unsigned long long cycle ) const
{
warp_inst_t inst_copy = inst;
@@ -1777,7 +1765,7 @@ public:
unsigned shader_id,
unsigned tpc_id,
const shader_core_config *config,
- const struct memory_config *mem_config,
+ const memory_config *mem_config,
shader_core_stats *stats );
// used by simt_core_cluster:
@@ -2072,7 +2060,7 @@ public:
simt_core_cluster( class gpgpu_sim *gpu,
unsigned cluster_id,
const shader_core_config *config,
- const struct memory_config *mem_config,
+ const memory_config *mem_config,
shader_core_stats *stats,
memory_stats_t *mstats );
generated by cgit v1.3 (git 2.53.0) at 2026-07-10 21:24:05 +0000