Author: Armin Rigo <[email protected]>
Branch: guard-compatible
Changeset: r94073:c4f28bd30a85
Date: 2018-03-22 15:14 +0100
http://bitbucket.org/pypy/pypy/changeset/c4f28bd30a85/
Log: hg merge vtune
diff --git a/rpython/jit/backend/x86/assembler.py
b/rpython/jit/backend/x86/assembler.py
--- a/rpython/jit/backend/x86/assembler.py
+++ b/rpython/jit/backend/x86/assembler.py
@@ -177,7 +177,8 @@
self.pop_gcmap(mc) # cancel the push_gcmap(store=True) in the caller
self._pop_all_regs_from_frame(mc, [], self.cpu.supports_floats)
mc.RET()
- self._frame_realloc_slowpath = mc.materialize(self.cpu, [])
+ self._frame_realloc_slowpath = self.materialize(mc, [],
+ "frame_realloc")
def _build_cond_call_slowpath(self, supports_floats, callee_only):
""" This builds a general call slowpath, for whatever call happens to
@@ -215,7 +216,7 @@
self._pop_all_regs_from_frame(mc, [eax], supports_floats, callee_only)
mc.RET()
self.flush_pending_slowpaths(mc)
- return mc.materialize(self.cpu, [])
+ return self.materialize(mc, [], "cond_call")
def _build_malloc_slowpath(self, kind):
""" While arriving on slowpath, we have a gcpattern on stack 0.
@@ -305,7 +306,7 @@
mc.JMP(imm(self.propagate_exception_path))
self.flush_pending_slowpaths(mc)
#
- rawstart = mc.materialize(self.cpu, [])
+ rawstart = self.materialize(mc, [], "malloc")
return rawstart
def _build_propagate_exception_path(self):
@@ -323,7 +324,7 @@
self.mc.MOV(RawEbpLoc(ofs), imm(propagate_exception_descr))
#
self._call_footer()
- rawstart = self.mc.materialize(self.cpu, [])
+ rawstart = self.materialize(self.mc, [], "propagate_exception")
self.propagate_exception_path = rawstart
self.mc = None
@@ -368,7 +369,7 @@
mc.ADD_ri(esp.value, WORD)
mc.JMP(imm(self.propagate_exception_path))
#
- rawstart = mc.materialize(self.cpu, [])
+ rawstart = self.materialize(mc, [], "stack_check")
self.stack_check_slowpath = rawstart
def _build_wb_slowpath(self, withcards, withfloats=False, for_frame=False):
@@ -469,7 +470,7 @@
mc.LEA_rs(esp.value, 7 * WORD)
mc.RET()
- rawstart = mc.materialize(self.cpu, [])
+ rawstart = self.materialize(mc, [], "write_barrier")
if for_frame:
self.wb_slowpath[4] = rawstart
else:
@@ -567,6 +568,7 @@
ops_offset=ops_offset)
self.fixup_target_tokens(rawstart)
+ self.materialize_done(rawstart, full_size, "loop%d" % looptoken.number)
self.teardown()
# oprofile support
if self.cpu.profile_agent is not None:
@@ -644,6 +646,8 @@
self.fixup_target_tokens(rawstart)
self.update_frame_depth(frame_depth)
+ self.materialize_done(rawstart, fullsize,
+ "loop%d" % original_loop_token.number)
self.teardown()
# oprofile support
if self.cpu.profile_agent is not None:
@@ -710,11 +714,14 @@
self.mc.JMP_l(0)
self.mc.writeimm32(0)
self.mc.force_frame_size(DEFAULT_FRAME_BYTES)
+ fullsize = self.mc.get_relative_pos()
rawstart = self.materialize_loop(looptoken)
# update the jump (above) to the real trace
self._patch_jump_to(rawstart + offset, asminfo.rawstart)
# update the guard to jump right to this custom piece of assembler
self.patch_jump_for_descr(faildescr, rawstart)
+ self.materialize_done(rawstart, fullsize,
+ "loop%d" % looptoken.number)
def _patch_jump_to(self, adr_jump_offset, adr_new_target):
assert adr_jump_offset != 0
@@ -892,13 +899,30 @@
mc.writeimm32(allocated_depth)
mc.copy_to_raw_memory(adr)
+ def get_asmmemmgr_blocks(self, looptoken):
+ clt = looptoken.compiled_loop_token
+ if clt.asmmemmgr_blocks is None:
+ clt.asmmemmgr_blocks = []
+ return clt.asmmemmgr_blocks
+
+ def materialize_done(self, rawstart, size, funcname):
+ from rpython.jit.backend.x86.vtune import rpy_vtune_register
+ with rffi.scoped_str2charp("rpyjit." + funcname) as p:
+ rpy_vtune_register(p, rawstart, size)
+
+ def materialize(self, mc, allblocks, funcname, gcrootmap=None):
+ size = mc.get_relative_pos()
+ rawstart = mc.materialize(self.cpu, allblocks, gcrootmap=gcrootmap)
+ self.materialize_done(rawstart, size, funcname)
+ return rawstart
+
def materialize_loop(self, looptoken):
self.datablockwrapper.done() # finish using cpu.asmmemmgr
self.datablockwrapper = None
allblocks = self.get_asmmemmgr_blocks(looptoken)
size = self.mc.get_relative_pos()
- res = self.mc.materialize(self.cpu, allblocks,
- self.cpu.gc_ll_descr.gcrootmap)
+ res = self.materialize(self.mc, allblocks,
+ gcrootmap=self.cpu.gc_ll_descr.gcrootmap)
if self.cpu.HAS_CODEMAP:
self.cpu.codemap.register_codemap(
self.codemap_builder.get_final_bytecode(res, size))
@@ -2072,7 +2096,7 @@
# now we return from the complete frame, which starts from
# _call_header_with_stack_check(). The _call_footer below does it.
self._call_footer()
- rawstart = mc.materialize(self.cpu, [])
+ rawstart = self.materialize(mc, [], "failure_recovery")
self.failure_recovery_code[exc + 2 * withfloats] = rawstart
self.mc = None
diff --git a/rpython/jit/backend/x86/guard_compat.py
b/rpython/jit/backend/x86/guard_compat.py
--- a/rpython/jit/backend/x86/guard_compat.py
+++ b/rpython/jit/backend/x86/guard_compat.py
@@ -288,7 +288,7 @@
mc.JMP_s(0)
assembler.flush_pending_slowpaths(mc)
- assembler.guard_compat_search_tree = mc.materialize(assembler.cpu, [])
+ assembler.guard_compat_search_tree = assembler.materialize(mc, [])
def build_once_guard_compat_recovery(assembler):
@@ -303,7 +303,7 @@
target = assembler.get_target_for_failure_recovery_of_guard_compat()
mc.JMP(regloc.imm(target))
- assembler.guard_compat_recovery = mc.materialize(assembler.cpu, [])
+ assembler.guard_compat_recovery = assembler.materialize(mc, [])
def generate_recovery_stub(assembler, guard_token):
diff --git a/rpython/jit/backend/x86/vtune.py b/rpython/jit/backend/x86/vtune.py
new file mode 100644
--- /dev/null
+++ b/rpython/jit/backend/x86/vtune.py
@@ -0,0 +1,43 @@
+"""
+Support for VTune Amplifier
+"""
+
+from rpython.rtyper.lltypesystem import lltype, rffi
+from rpython.translator.tool.cbuild import ExternalCompilationInfo
+
+
+eci = ExternalCompilationInfo(
+ post_include_bits=["""
+RPY_EXTERN void rpy_vtune_register(char *, long, long);
+"""],
+ include_dirs=["/opt/intel/system_studio_2018/vtune_amplifier/include"],
+ libraries=["dl"], # otherwise, iJIT_IsProfilingActive() just returns 0
+ separate_module_sources=["""
+#include
"/opt/intel/system_studio_2018/vtune_amplifier/sdk/src/ittnotify/jitprofiling.c"
+
+RPY_EXTERN void rpy_vtune_register(char *funcname, Signed addr, Signed size)
+{
+ iJIT_Method_Load_V2 jmethod = {0};
+
+ if (iJIT_IsProfilingActive() != iJIT_SAMPLING_ON) {
+ return;
+ }
+
+ jmethod.method_id = iJIT_GetNewMethodID();
+ jmethod.method_name = funcname;
+ jmethod.method_load_address = (void *)addr;
+ jmethod.method_size = size;
+ jmethod.module_name = "rpyjit";
+
+ iJIT_NotifyEvent(iJVM_EVENT_TYPE_METHOD_LOAD_FINISHED_V2,
+ (void*)&jmethod);
+}
+"""])
+
+rpy_vtune_register = rffi.llexternal(
+ "rpy_vtune_register",
+ [rffi.CCHARP, lltype.Signed, lltype.Signed],
+ lltype.Void,
+ compilation_info=eci,
+ _nowrapper=True,
+ sandboxsafe=True)
_______________________________________________
pypy-commit mailing list
[email protected]
https://mail.python.org/mailman/listinfo/pypy-commit
