Fix 32bit wrap to att buffer size (#176)

Co-authored-by: Giovanni Baraldi <gbaraldi@amd.com>

Everything but PSDB passed. I'm force merging because 1) this is a minor change and 2) PSDBs seem broken now, they are stuck and pointing to the wrong location: http://rocm-ci.amd.com/job/compute-psdb-staging-profiler-emu/435/

Author: Baraldi, Giovanni

gfxip/gfx10/gfx10_primitives.h

@@ -618,7 +618,7 @@ class gfx10_cntx_prim {
 
   // Indicates the size of buffer to use per Shader Engine instance.
   // The size is specified in terms of 4KB blocks
-  static uint32_t sqtt_buffer_size_value(uint32_t size_val, uint32_t base_hi) {
+  static uint32_t sqtt_buffer_size_value(uint64_t size_val, uint32_t base_hi) {
 #if SQTT_PRIM_ENABLED
     uint32_t size{0};
     size = SET_REG_FIELD_BITS(SQ_THREAD_TRACE_BUF0_SIZE, SIZE, size_val >> TT_BUFF_ALIGN_SHIFT) |

gfxip/gfx11/gfx11_primitives.h

@@ -639,7 +639,7 @@ class gfx11_cntx_prim {
 
   // Indicates the size of buffer to use per Shader Engine instance.
   // The size is specified in terms of 4KB blocks
-  static uint32_t sqtt_buffer_size_value(uint32_t size_val, uint32_t base_hi) {
+  static uint32_t sqtt_buffer_size_value(uint64_t size_val, uint32_t base_hi) {
 #if SQTT_PRIM_ENABLED
     uint32_t sq_thread_trace_buf0_size =
         SET_REG_FIELD_BITS(SQ_THREAD_TRACE_BUF0_SIZE, SIZE, size_val >> TT_BUFF_ALIGN_SHIFT) |

gfxip/gfx12/gfx12_primitives.h

@@ -560,7 +560,7 @@ class gfx12_cntx_prim {
 
   // Indicates the size of buffer to use per Shader Engine instance.
   // The size is specified in terms of 4KB blocks
-  static uint32_t sqtt_buffer0_size_value(uint32_t size_val) {
+  static uint32_t sqtt_buffer0_size_value(uint64_t size_val) {
     uint32_t sq_thread_trace_buf0_size{0};
     sq_thread_trace_buf0_size =
         SET_REG_FIELD_BITS(SQ_THREAD_TRACE_BUF0_SIZE, SIZE, size_val >> TT_BUFF_ALIGN_SHIFT);

gfxip/gfx9/gfx9_primitives.h

@@ -683,7 +683,7 @@ class gfx9_cntx_prim {
 
   // Indicates the size of buffer to use per Shader Engine instance.
   // The size is specified in terms of 4KB blocks
-  static uint32_t sqtt_buffer_size_value(uint32_t size_val, uint32_t base_hi) {
+  static uint32_t sqtt_buffer_size_value(uint64_t size_val, uint32_t base_hi) {
     uint32_t sq_thread_trace_size =
         SET_REG_FIELD_BITS(SQ_THREAD_TRACE_SIZE, SIZE, (size_val >> TT_BUFF_ALIGN_SHIFT));
     return sq_thread_trace_size;