tinygrad/tinygrad
2
0
Fork 3
mirror of https://github.com/tinygrad/tinygrad.git synced 2026-08-29 19:36:07 +00:00
Code Issues Packages Projects Releases Wiki Activity

minor patches for SQTT over usb on gfx12 (#12627)

Browse Source 
* disable cpu_access in the sqtt buffer allocation

not sure if this is required, it results in a very slow call to
pcie_mem_write over USB GPU, removing it worked fine.

* fix itrace_se_mask on gfx12

on gfx11 it gave 6 se, on gfx11 this value is 2 so no instructions were
traced.

* Revert "fix itrace_se_mask on gfx12"

This reverts commit 0644adbcd1.
This commit is contained in:
qazal
2025-10-14 18:07:46 +08:00
committed by GitHub
parent 1e6e5a0efd
commit d3bfcd3277
1 changed files with 1 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tinygrad/runtime/ops_amd.py
+1 -1
View File
@@ -819,7 +819,7 @@ class AMDDevice(HCQCompiled):
f"ppfeaturemask={(ppfeaturemask&~0x8000):#x} (current {ppfeaturemask=:#x} & ~PP_GFXOFF_MASK) to amdgpu module parameters\n"
"For more information read https://github.com/tinygrad/tinygrad/blob/master/extra/sqtt/README.md")
SQTT_BUFFER_SIZE = getenv("SQTT_BUFFER_SIZE", 256) # in mb, per shader engine
self.sqtt_buffers = [self.allocator.alloc(SQTT_BUFFER_SIZE*1024*1024, BufferSpec(cpu_access=True, nolru=True)) for _ in range(self.se_cnt)]
self.sqtt_buffers = [self.allocator.alloc(SQTT_BUFFER_SIZE*1024*1024, BufferSpec(nolru=True)) for _ in range(self.se_cnt)]
self.sqtt_itrace_se_mask = getenv("SQTT_ITRACE_SE_MASK", 2) # -1 enable all, 0 disable all, >0 bitmask for where to enable instruction tracing
self.sqtt_next_cmd_id = itertools.count(0)
cast(AMDComputeQueue, self.hw_compute_queue_t()).sqtt_start(self.sqtt_buffers, self.sqtt_itrace_se_mask).submit(self)