From 0c716e4c6b7c2c44bbfd02b6503cae66af7b7480 Mon Sep 17 00:00:00 2001 From: Marshall Lochbaum Date: Fri, 28 Jan 2022 16:34:41 -0500 Subject: Separate syntax highlighting category for header/body characters ;:? --- docs/implementation/kclaims.html | 48 ++++++++++++++++++++-------------------- docs/implementation/vm.html | 6 ++--- 2 files changed, 27 insertions(+), 27 deletions(-) (limited to 'docs/implementation') diff --git a/docs/implementation/kclaims.html b/docs/implementation/kclaims.html index 26bbe7ca..d99d8ea4 100644 --- a/docs/implementation/kclaims.html +++ b/docs/implementation/kclaims.html @@ -41,35 +41,35 @@ [Cycles where a code fetch is stalled due to L1 instruction cache miss]

That's just the whole cost (in cycles) of L1 misses, exactly what we want! First I'll run this on a J program I have lying around, building my old Honors thesis with JtoLaTeX.

-
 Performance counter stats for 'jlatex document.jtex nopdf':
+
 Performance counter stats for 'jlatex document.jtex nopdf':
 
-     1,457,284,402      cycles:u
-        56,485,452      icache_16b.ifdata_stall:u
-         2,254,192      cache-misses:u
-        37,849,426      L1-dcache-load-misses:u
-        28,797,332      L1-icache-load-misses:u
+     1,457,284,402      cycles:u
+        56,485,452      icache_16b.ifdata_stall:u
+         2,254,192      cache-misses:u
+        37,849,426      L1-dcache-load-misses:u
+        28,797,332      L1-icache-load-misses:u
 
        0.557255985 seconds time elapsed
 

 
Here's the BQN call that builds CBQN's object code sources:

-
 Performance counter stats for './genRuntime /home/marshall/BQN/':
+
 Performance counter stats for './genRuntime /home/marshall/BQN/':
 
-       241,224,322      cycles:u
-         5,452,372      icache_16b.ifdata_stall:u
-           829,146      cache-misses:u
-         6,954,143      L1-dcache-load-misses:u
-         1,291,804      L1-icache-load-misses:u
+       241,224,322      cycles:u
+         5,452,372      icache_16b.ifdata_stall:u
+           829,146      cache-misses:u
+         6,954,143      L1-dcache-load-misses:u
+         1,291,804      L1-icache-load-misses:u
 
        0.098228740 seconds time elapsed
 

 
And the Python-based font tool I use to build font samples for this site:

-
 Performance counter stats for 'pyftsubset […more stuff]':
+
 Performance counter stats for 'pyftsubset […more stuff]':
 
-       499,025,775      cycles:u
-        24,869,974      icache_16b.ifdata_stall:u
-         5,850,063      cache-misses:u
-        11,175,902      L1-dcache-load-misses:u
-        11,784,702      L1-icache-load-misses:u
+       499,025,775      cycles:u
+        24,869,974      icache_16b.ifdata_stall:u
+         5,850,063      cache-misses:u
+        11,175,902      L1-dcache-load-misses:u
+        11,784,702      L1-icache-load-misses:u
 
        0.215698059 seconds time elapsed
 

@@ -84,13 +84,13 @@
 
So, roughly 4%, 2%, and 5%. The cache miss counts are also broadly in line with these numbers. Note that full cache misses are pretty rare, so that most misses just hit L2 or L3 and don't suffer a large penalty. Also note that instruction cache misses are mostly lower than data misses, as expected.

 
Don't get me wrong, I'd love to improve performance even by 2%. But it's not exactly world domination, is it? And it doesn't matter how cache-friendly K is, that's the absolute limit.

 
For comparison, here's ngn/k (which does aim for a small executable) running one of its unit tests—test 19 in the a20/ folder, chosen because it's the longest-running of those tests.

-
 Performance counter stats for '../k 19.k':
+
 Performance counter stats for '../k 19.k':
 
-     3,341,989,998      cycles:u
-        21,136,960      icache_16b.ifdata_stall:u
-           336,847      cache-misses:u
-        10,748,990      L1-dcache-load-misses:u
-        20,204,548      L1-icache-load-misses:u
+     3,341,989,998      cycles:u
+        21,136,960      icache_16b.ifdata_stall:u
+           336,847      cache-misses:u
+        10,748,990      L1-dcache-load-misses:u
+        20,204,548      L1-icache-load-misses:u
 
        1.245378356 seconds time elapsed
 

diff --git a/docs/implementation/vm.html b/docs/implementation/vm.html
index 41bfee29..7a9db034 100644
--- a/docs/implementation/vm.html
+++ b/docs/implementation/vm.html
@@ -33,7 +33,7 @@
 
The last property can be a single number or a list of lists. A single number indicates the body to be executed, and is used only for blocks with exactly one body. If it's a list of lists, the length is 1 for a block without arguments and 2 or more for a block with arguments (function or deferred modifier). Each element is a list of body indices. After selecting the appropriate list, execution begins at the first body in the appropriate list, moving to the next one if a header test (SETH or PRED instruction) fails. If a test fails but there's no next body, block evaluation is an error.

 
The five possible cases for a function are monadic, dyadic, inverse monadic (𝕊⁼x), inverse dyadic (w𝕊⁼x), and swapped-inverse dyadic (x𝕊˜⁼w). The first two will always be provided, while the remaining three typically don't exist as they have to be specified with undo headers. The smallest length that covers all possible cases will be used.

 
Bodies

-
Bodies in a block are separated by ;. Each entry in bodies is a list containing:

+
Bodies in a block are separated by ;. Each entry in bodies is a list containing:

 
    
 
  • Starting index in code
    • 
 
  • Number of variables the block needs to allocate
    • 
@@ -43,7 +43,7 @@
 
    The starting index refers to the position in bytecode where execution starts in order to evaluate the block. Different bodies will always have the same set of special names, but the variables they define are unrelated, so of course they can have different counts. The given number of variables includes special names, but list of names and export mask don't.
    
 
    The program's symbol list is included in the tokenization information t: it is 0⊑2⊑t. Since the entire program (the source code passed in one compiler call) uses this list, namespace field accesses can be performed with indices alone within a program. The symbol list is needed for cross-program access, for example if •BQN returns a namespace.
    
 
    Instructions
    
-
    The following instructions are defined (those without names are tentatively reserved only). The ones emitted by the self-hosted BQN compiler are marked in the "used" column. Instructions marked "NS" are used only in programs with namespaces, and those marked "HE" are used only with headers : or predicates ?. Only those marked "X" are needed to support the compiler and self-hosted runtime.
    
+
    The following instructions are defined (those without names are tentatively reserved only). The ones emitted by the self-hosted BQN compiler are marked in the "used" column. Instructions marked "NS" are used only in programs with namespaces, and those marked "HE" are used only with headers : or predicates ?. Only those marked "X" are needed to support the compiler and self-hosted runtime.
    
 
@@ -400,7 +400,7 @@
 
-
+
-- 
cgit v1.2.3


    
 
 
    
 08
 RETDx? → nx? → n
 Clears stack, dropping 0 or 1 value