From afae11ea884b023f74dbbc773f5de7d41afde29c Mon Sep 17 00:00:00 2001 From: Raymond Hettinger Date: Sat, 5 Jul 2008 02:11:55 +0000 Subject: [PATCH] Update comment on prediction macros. --- Python/ceval.c | 22 ++++++++++++---------- 1 file changed, 12 insertions(+), 10 deletions(-) diff --git a/Python/ceval.c b/Python/ceval.c index 037c86c134a..a9e37ae1fcb 100644 --- a/Python/ceval.c +++ b/Python/ceval.c @@ -615,18 +615,20 @@ PyEval_EvalFrameEx(PyFrameObject *f, int throwflag) COMPARE_OP is often followed by JUMP_IF_FALSE or JUMP_IF_TRUE. And, those opcodes are often followed by a POP_TOP. - Verifying the prediction costs a single high-speed test of register + Verifying the prediction costs a single high-speed test of a register variable against a constant. If the pairing was good, then the - processor has a high likelihood of making its own successful branch - prediction which results in a nearly zero overhead transition to the - next opcode. + processor's own internal branch predication has a high likelihood of + success, resulting in a nearly zero-overhead transition to the + next opcode. A successful prediction saves a trip through the eval-loop + including its two unpredictable branches, the HAS_ARG test and the + switch-case. Combined with the processor's internal branch prediction, + a successful PREDICT has the effect of making the two opcodes run as if + they were a single new opcode with the bodies combined. - A successful prediction saves a trip through the eval-loop including - its two unpredictable branches, the HAS_ARG test and the switch-case. - - If collecting opcode statistics, turn off prediction so that - statistics are accurately maintained (the predictions bypass - the opcode frequency counter updates). + If collecting opcode statistics, your choices are to either keep the + predictions turned-on and interpret the results as if some opcodes + had been combined or turn-off predictions so that the opcode frequency + counter updates for both opcodes. */ #ifdef DYNAMIC_EXECUTION_PROFILE