From f103f929522f964501de1a7f6935fc3b294d5790 Mon Sep 17 00:00:00 2001
From: Inderpreet Singh <w6o6@ece.ubc.ca>
Date: Fri, 2 Mar 2012 03:39:08 -0800
Subject: Changed arch_rech type to store 16 registers, 8 input and 8 output. 8
 inputs because one can have 4 input operands and 4 register operands in a
 surface store instruction. Fixed arch_regs for memory instructions being
 ignored in the pre-decode statge.

[git-p4: depot-paths = "//depot/gpgpu_sim_research/fermi/distribution/": change = 11576]
---
 src/gpgpu-sim/shader.cc | 12 ++++++------
 src/gpgpu-sim/shader.h  |  6 +++---
 2 files changed, 9 insertions(+), 9 deletions(-)

(limited to 'src/gpgpu-sim')

diff --git a/src/gpgpu-sim/shader.cc b/src/gpgpu-sim/shader.cc
index 060dd07..82be8ed 100644
--- a/src/gpgpu-sim/shader.cc
+++ b/src/gpgpu-sim/shader.cc
@@ -54,8 +54,8 @@
 std::list<unsigned> shader_core_ctx::get_regs_written( const inst_t &fvt ) const
 {
    std::list<unsigned> result;
-   for( unsigned op=0; op < 4; op++ ) {
-      int reg_num = fvt.arch_reg[op]; // this math needs to match that used in function_info::ptx_decode_inst
+   for( unsigned op=0; op < MAX_REG_OPERANDS; op++ ) {
+      int reg_num = fvt.arch_reg.dst[op]; // this math needs to match that used in function_info::ptx_decode_inst
       if( reg_num >= 0 ) // valid register
          result.push_back(reg_num);
    }
@@ -1976,7 +1976,7 @@ void opndcoll_rfu_t::collector_unit_t::dump(FILE *fp, const shader_core_ctx *sha
       fprintf(fp,"    <free>\n");
    } else {
       m_warp->print(fp);
-      for( unsigned i=0; i < MAX_REG_OPERANDS; i++ ) {
+      for( unsigned i=0; i < MAX_REG_OPERANDS*2; i++ ) {
          if( m_not_ready.test(i) ) {
             std::string r = m_src_op[i].get_reg_string();
             fprintf(fp,"    '%s' not ready\n", r.c_str() );
@@ -2007,8 +2007,8 @@ void opndcoll_rfu_t::collector_unit_t::allocate( warp_inst_t** pipeline_reg, war
    m_output_register = output_reg;
    if( !(*pipeline_reg)->empty() ) {
       m_warp_id = (*pipeline_reg)->warp_id();
-      for( unsigned op=0; op < 4; op++ ) {
-         int reg_num = (*pipeline_reg)->arch_reg[4+op]; // this math needs to match that used in function_info::ptx_decode_inst
+      for( unsigned op=0; op < MAX_REG_OPERANDS; op++ ) {
+         int reg_num = (*pipeline_reg)->arch_reg.src[op]; // this math needs to match that used in function_info::ptx_decode_inst
          if( reg_num >= 0 ) { // valid register
             m_src_op[op] = op_t( this, op, reg_num, m_num_banks, m_bank_warp_shift );
             m_not_ready.set(op);
@@ -2025,7 +2025,7 @@ void opndcoll_rfu_t::collector_unit_t::dispatch()
    move_warp(*m_output_register,m_warp);
    m_free=true;
    m_output_register = NULL;
-   for( unsigned i=0; i<MAX_REG_OPERANDS;i++) 
+   for( unsigned i=0; i<MAX_REG_OPERANDS*2;i++)
       m_src_op[i].reset();
 }
 
diff --git a/src/gpgpu-sim/shader.h b/src/gpgpu-sim/shader.h
index f8b0e58..b0a9c3a 100644
--- a/src/gpgpu-sim/shader.h
+++ b/src/gpgpu-sim/shader.h
@@ -490,7 +490,7 @@ private:
       void add_read_requests( collector_unit_t *cu ) 
       {
          const op_t *src = cu->get_operands();
-         for( unsigned i=0; i<MAX_REG_OPERANDS; i++) {
+         for( unsigned i=0; i<MAX_REG_OPERANDS*2; i++) {
             const op_t &op = src[i];
             if( op.valid() ) {
                unsigned bank = op.get_bank();
@@ -554,7 +554,7 @@ private:
          m_free = true;
          m_warp = NULL;
          m_output_register = NULL;
-         m_src_op = new op_t[MAX_REG_OPERANDS];
+         m_src_op = new op_t[MAX_REG_OPERANDS*2];
          m_not_ready.reset();
          m_warp_id = -1;
          m_num_banks = 0;
@@ -591,7 +591,7 @@ private:
       warp_inst_t  *m_warp;
       warp_inst_t** m_output_register; // pipeline register to issue to when ready
       op_t *m_src_op;
-      std::bitset<MAX_REG_OPERANDS> m_not_ready;
+      std::bitset<MAX_REG_OPERANDS*2> m_not_ready;
       unsigned m_num_banks;
       unsigned m_bank_warp_shift;
       opndcoll_rfu_t *m_rfu;
-- 
cgit v1.3