Here is the information which I have parsed out of an avcC atom in an mp4 container
The avc extradata Conforms with ISO/IEC 14496-15:2004(E) 5.2.4.1.1
> 0x01 0x42 0x00 0x1E 0xFF 0xE1 0x00 0x0E
Configuration Version: 1 u(8)<br>
AVCProfileIndication: 66 u(8)<br>
profile_compatability: 0 u(8)<br>
AVCLevelIndication: 30 u(8)<br>
bit(6) reserved = '111111' b <br>
unsigned int (2) lengthSizeMinusOne = '11' <br>
bit(3) reseved = '111' <br>
unsigned int (5) numOfSequenceParameterSets = 1 <br>
unsigned int (16) sequenceParameterSetLength = 14 <br>
SPS
> 0x67 0x42 0x00 0x1E 0x8D 0x68 0x6E 0x03 0xDA 0x6A 0x0C 0x02 0x0C 0x04 <br>
avC data Continued
> 0x01 0x00 0x04 <br>
unsigned int (8) numOfPictureParameterSets: 1 <br>
unsigned int (16) pictureParameterSetLength: 4 <br>
PPS <br>
>0x68 0xCE 0x74 0xC8
The contents of the SPS appears to give incorrect results regarding pic_width_in_mbs_minus1 (5) and I do not believe there are any emulation_3_byte preventions. Am I missing something obvious? I am parsing the SPS according to ISO/IEC 14496-10:2004(E) which is the same SPS parsing information found here.
Image is 96x16 (why?)
Sequence Parameter Set
profile_idc 66
constraint_set0_flag 0
constraint_set1_flag 0
constraint_set2_flag 0
constraint_set3_flag 0
level_idc 30
seq_parameter_set_id 0
// ...
num_ref_frames 1
gaps_in_frame_num_value_allowed_flag 0
pic_width_in_mbs_minus1 5
pic_height_in_map_units_minus1 0
frame_mbs_only_flag 1
direct_8x8_inference_flag 1
frame_cropping_flag 0
vui_parameters_present_flag 0
// ...
Picture Parameter Set
pic_parameter_set_id 0
seq_parameter_set_id 0
entropy_coding_mode_flag 0
num_slice_groups_minus1 0
// ...
Related
I'm relatively new to lua and programming in general (self taught), so please be gentle!
Anyway, I wrote a lua script to read a UDP message from a game. The structure of the message is:
DATAxXXXXaaaaBBBBccccDDDDeeeeFFFFggggHHHH
DATAx = 4 letter ID and x = control character
XXXX = integer shows the group of the data (groups are known)
aaaa...HHHHH = 8 single-precision floating point numbers
The last ones is those numbers I need to decode.
If I print the message as received, it's something like:
DATA*{V???A?A?...etc.
Using string.byte(), I'm getting a stream of bytes like this (I have "formatted" the bytes to reflect the structure above.
68 65 84 65/42/20 0 0 0/237 222 28 66/189 59 182 65/107 42 41 65/33 173 79 63/0 0 128 63/146 41 41 65/0 0 30 66/0 0 184 65
The first 5 bytes are of course the DATA*. The next 4 are the 20th group of data. The next bytes, the ones I need to decode, and are equal to those values:
237 222 28 66 = 39.218
189 59 182 65 = 22.779
107 42 41 65 = 10.573
33 173 79 63 = 0.8114
0 0 128 63 = 1.0000
146 41 41 65 = 10.573
0 0 30 66 = 39.500
0 0 184 65 = 23.000
I've found C# code that does the decode with BitConverter.ToSingle(), but I haven't found any like this for Lua.
Any idea?
What Lua version do you have?
This code works in Lua 5.3
local str = "DATA*\20\0\0\0\237\222\28\66\189\59\182\65..."
-- Read two float values starting from position 10 in the string
print(string.unpack("<ff", str, 10)) --> 39.217700958252 22.779169082642 18
-- 18 (third returned value) is the next position in the string
For Lua 5.1 you have to write special function (or steal it from François Perrad's git repo )
local function binary_to_float(str, pos)
local b1, b2, b3, b4 = str:byte(pos, pos+3)
local sign = b4 > 0x7F and -1 or 1
local expo = (b4 % 0x80) * 2 + math.floor(b3 / 0x80)
local mant = ((b3 % 0x80) * 0x100 + b2) * 0x100 + b1
local n
if mant + expo == 0 then
n = sign * 0.0
elseif expo == 0xFF then
n = (mant == 0 and sign or 0) / 0
else
n = sign * (1 + mant / 0x800000) * 2.0^(expo - 0x7F)
end
return n
end
local str = "DATA*\20\0\0\0\237\222\28\66\189\59\182\65..."
print(binary_to_float(str, 10)) --> 39.217700958252
print(binary_to_float(str, 14)) --> 22.779169082642
It’s little-endian byte-order of IEEE-754 single-precision binary:
E.g., 0 0 128 63 is:
00111111 10000000 00000000 00000000
(63) (128) (0) (0)
Why that equals 1 requires that you understand the very basics of IEEE-754 representation, namely its use of an exponent and mantissa. See here to start.
See #Egor‘s answer above for how to use string.unpack() in Lua 5.3 and one possible implementation you could use in earlier versions.
I need to realise a PWM output with 5 kHz +/- 5%. (Presumably due to a filter circuit into which this feeds over which I have no control.)
Is this realisable with a ESP8266 (ideally with NodeMCU)?
I realise that the software PWM of the ESP8266 has a maximum frequency of 1 kHz while the sigma-delta can be used to implement a PWM with a fixed frequency of about 300 kHz.
So is there a reliable way to achieve 5 kHz? I know some people experimented with the I2S peripheral for waveform output but I am unsure whether it can be used for 5kHz output.
Has anybody looked at into a similar problem before?
The essential code is:
#define qap 2 // Quick As Possible ... Duty cycle only 0, 50, 100%
#define HFreq 5150
#define pPulse D2 // a NodeMCU/ESP8266 GPIO PWM pin
analogWriteRange(qap); analogWriteFreq( HFreq ); analogWrite(pPulse, 1); // start PWM
TL;DR
Just did some crude benchamrks
// had HFreq=126400 with 75 KHz pulse at 80 MHz ESP clock, every 1 sec but proof of evidence lost
nominally with 80 MHz clock got
// 72.24 KHz 361178 2015061929
// 72.23 KHz 361163 2415062390
// 72.23 KHz 361133 2815062824
and
// 141.52 KHz 353809 2009395076
// 141.54 KHz 353846 2409395627
// 141.52 KHz 353806 2809395946
with 160 MHz clock
CAVEATS!!!
1. The duty cycle range is 2!
2. As well the system had no other conditioning, so other timing artifacts were still present from Serial IO etc. In particular instability due to the WDT and WiFi penultimately appeared. (Anyhow, presumably this is an issue only when the ESP8266 is under stress and duress.)
// 141.50 KHz 353754 619466806
// 141.52 KHz 353810 1019467038
// ...
// ad infinitum cum tempore finitum, infinitus est ad nauseum?
// ...
// 141.54 KHz 353857 735996888
//
// ets Jan 8 2013,rst cause:4, boot mode:(1,7)
//
//wdt reset
When using the following code to generate a 5 KHz square wave signal, the considerations above are not an issue and do not occur.
// ------------ test results for 80 MHz clock --------------
//
//
// PWM pulse test
//
// F_CPU: 80000000L
// ESP8266_CLOCK: 80000000UL
// PWM "freq.": 5150
//
//
// connect D1 to D2
//
//
// raw MPU
// frequency count cycle
// 0.00 KHz 1 407976267
// 4.74 KHz 9482 567976702
// 5.00 KHz 10007 727977137
// 5.00 KHz 10006 887977572
// 5.00 KHz 10006 1047978007
// 5.00 KHz 10007 1207978442
// 5.00 KHz 10006 1367978877
// 5.00 KHz 10006 1527979312
// 5.00 KHz 10007 1687979747
// 5.00 KHz 10006 1847980182
// 5.00 KHz 10006 2007980617
// 5.00 KHz 10007 2167981052
// 5.00 KHz 10006 2327981487
// 5.00 KHz 10006 2487981922
// 5.00 KHz 10007 2647982357 ...
//
// crude testing for 5KHz signal
// extracted from:
// highest frequency / shortest period pin pulse generate / detect test
//
// uses raw ESP8266 / NodeMCU V1.0 hardware primitive interface of Arduino IDE (no included libraries)
//
// timing dependencies: WDT, WiFi, I2S, I2C, one wire, UART, SPI, ...
//
// Arduino GPIO 16 5 4 0 2 14 12 13 15 3 1 0 1 2 3 4 5 ... 12 13 14 15 16
// NodeMCU D pin 0 1 2 3 4 5 6 7 8 9 10 3 10 4 9 2 1 ... 6 7 5 8 0
// | | | | | | | | | | |
// a WAKE | | F Tx1 | | Rx2 Tx2 Rx0 Tx0
// k (NO PWM or | | L blue | | | |
// a' interrupt) | S A * H | H | H | | * led's
// s red S D S S M S H
// * C A H C I I C
// L T L S M S
// K A K O O
// └ - - - - └----UART's----┘
// └--I2C--┘ └-----SPI------┘
//
// rules of engagement are obscure and vague for effects of argument values for the paramters of these functions:
// analogWriteRange(qap); analogWriteFreq( HFreq ); analogWrite(pPulse, 1);
//
// http://stackoverflow.com/questions/42112357/how-to-implement-esp8266-5-khz-pwm
//
// system #defines: F_CPU ESP8266_CLOCK
#define pInt D1 // HWI pin: NOT D0 ie. GPIO16 is not hardwared interrupt or PWM pin
#define pPulse D2 // PWM pulsed frequency source ... ditto D0 (note: D4 = blue LED)
#define countFor 160000000UL
#define gmv(p) #p // get macro value
#define em(p) gmv(p) // evaluate macro
#define qap 2 // minimal number of duty cycle levels (0, 50, 100% ) Quick As Possible ...
#define HFreq 5150 //((long int) F_CPU==80000000L ? 125000 : 250000) // ... to minimize time of a cycle period
// max values ^ and ^ found empirically
// had HFreq=126400 with 75 KHz pulse at 80 MHz ESP clock, every 1 sec but proof of evidence lost
#define infoTxt (String) \
"\n\n\t PWM pulse test " \
"\n F_CPU: " em(F_CPU) \
"\n ESP8266_CLOCK: " em(ESP8266_CLOCK) \
"\n PWM \"freq.\": " + HFreq + "\n" \
"\n\n connect " em(pInt) " to " em(pPulse) "\n" \
"\n\n raw MPU " \
" \n frequency count cycle "
long int oc=1, cntr=1;
unsigned long int tc=0;
void hwISR(){ cntr++; } // can count pulses if pInt <---> to pPulse
void anISR(){ tc=ESP.getCycleCount(); timer0_write( countFor + tc ); oc=cntr; cntr=1; }
void setup() { // need to still confirm duty cycle=50% (scope it or ...)
noInterrupts();
Serial.begin(115200); Serial.println(infoTxt); delay(10); // Serial.flush(); Serial.end(); // Serial timing?
analogWriteRange(qap); analogWriteFreq( HFreq ); analogWrite(pPulse, 1); // start PWM
pinMode( pInt, INPUT ); attachInterrupt(pInt, hwISR, RISING); // count pulses
timer0_isr_init(); timer0_attachInterrupt( anISR ); anISR(); //
interrupts();
}
void loop() { delay(10); if (oc==0) return;
Serial.println((String)" "+(oc/1000.0*F_CPU/countFor)+" KHz "+oc+" "+tc); oc=0; }
//
You can also use the hardware SPI interface which has an adjustable clock speed.
By writing continuous data, the output waveform appears on SCLK.
I've figured out the abovementioned while creating a knight rider effect with IC 74hc595 on ESP8266, using MicroPython (which is indeed slow, as being a script language).
It worked for me up to 4MHz SPI clock speed.
The disadvantage is, for permanent waveform, you need to write data to MOSI forever (as when SPI data buffer goes empty, there is no signal on SCLK anymore).
I'm writing an simulator (for learning purposes) for complete M3UA-SCCP-TCAP-MAP stack (over SCTP). So far M3UA+SCCP stacks are OK.
M3UA Based on the RFC 4666 Sept 2006
SCCP Based on the ITU-T Q.711-Q716
TCAP Based on the ITU-T Q.771-Q775
But upon decoding TCAP part I got lost on dialoguePortion.
TCAP is asn.1 encoded, so everything is tag+len+data.
Wireshark decode it differently than my decoder.
Message is:
62434804102f00676b1e281c060700118605010101a011600f80020780a1090607040000010005036c1ba1190201010201163011800590896734f283010086059062859107
Basically, my message is BER-decoded as
Note: Format: hex(tag) + (BER splitted to CLS+PC+TAG in decimal) + hex(data)
62 ( 64 32 2 )
48 ( 64 0 8 ) 102f0067
6b ( 64 32 11 )
28 ( 0 32 8 )
06 ( 0 0 6 ) 00118605010101 OID=0.0.17.773.1.1.1
a0 ( 128 32 0 )
60 ( 64 32 0 )
80 ( 128 0 0 ) 0780
a1 ( 128 32 1 )
06 ( 0 0 6 ) 04000001000503 OID=0.4.0.0.1.0.5.3
6c ( 64 32 12 )
...
So I can see begin[2] message containing otid[8], dialogPortion[11] and componentPortion[12].
otid and ComponentPortion are decoded correctly. But not dialogPortion.
ASN for dialogPortion does not mention any of these codes.
Even more confusing, wireshark decode it differently (oid-as-dialogue is NOT in the dialoguePortion, but as a field after otid, which is NOT as described in ITU-T documentation - or not as I'm understanding it)
Wireshark decoded Transaction Capabilities Application Part
begin
Source Transaction ID
otid: 102f0067
oid: 0.0.17.773.1.1.1 (id-as-dialogue)
dialogueRequest
Padding: 7
protocol-version: 80 (version1)
1... .... = version1: True
application-context-name: 0.4.0.0.1.0.5.3 (locationInfoRetrievalContext-v3)
components: 1 item
...
I can't find any reference for Padding in dialoguePDU ASN.
Can someone point me in the right direction?
I would like to know how to properly decode this message
DialoguePDU format should be simple in this case:
dialogue-as-id OBJECT IDENTIFIER ::= {itu-t recommendation q 773 as(1) dialogue-as(1) version1(1)}
DialoguePDU ::= CHOICE {
dialogueRequest AARQ-apdu,
dialogueResponse AARE-apdu,
dialogueAbort ABRT-apdu
}
AARQ-apdu ::= [APPLICATION 0] IMPLICIT SEQUENCE {
protocol-version [0] IMPLICIT BIT STRING {version1(0)} DEFAULT {version1},
application-context-name [1] OBJECT IDENTIFIER,
user-information [30] IMPLICIT SEQUENCE OF EXTERNAL OPTIONAL
}
Wireshark is still wrong :-). But then... that is display. It displays values correctly - only in the wrong section. Probably some reason due to easier decoding.
What I was missing was definition of EXTERNAL[8]. DialoguePortion is declared as EXTERNAL...so now everything makes sense.
For your message, my very own decoder says:
begin [APPLICATION 2] (x67)
otid [APPLICATION 8] (x4) =102f0067h
dialoguePortion [APPLICATION 11] (x30)
EXTERNAL (x28)
direct-reference [OBJECT IDENTIFIER] (x7) =00118605010101h
encoding:single-ASN1-type [0] (x17)
dialogueRequest [APPLICATION 0] (x15)
protocol-version [0] (x2) = 80 {version1 (0) } spare bits= 7
application-context-name [1] (x9)
OBJECT IDENTIFIER (x7) =04000001000503h
components [APPLICATION 12] (x27)
invoke [1] (x25)
invokeID [INTEGER] (x1) =1d (01h)
operationCode [INTEGER] (x1) = (22) SendRoutingInfo
parameter [SEQUENCE] (x17)
msisdn [0] (x5) = 90896734f2h
Nature of Address: international number (1)
Numbering Plan Indicator: unknown (0)
signal= 9876432
interrogationType [3] (x1) = (0) basicCall
gmsc-Address [6] (x5) = 9062859107h
Nature of Address: international number (1)
Numbering Plan Indicator: unknown (0)
signal= 26581970
Now, wireshark's padding 7 and my spare bits=7 both refer to the the protocol-version field, defined in Q.773 as:
AARQ-apdu ::= [APPLICATION 0] IMPLICIT SEQUENCE {
protocol-version [0] IMPLICIT BIT STRING { version1 (0) }
DEFAULT { version1 },
application-context-name [1] OBJECT IDENTIFIER,
user-information [30] IMPLICIT SEQUENCE OF EXTERNAL
OPTIONAL }
the BIT STRING definition, assigns name to just the leading bit (version1)... the rest (7 bits) are not given a name and wireshark consider them as padding
I created NSData of length 2 bytes (16 bits) and I want to set first 12 bits as binary value of (int)120 and 13th bit as 0 or 1(bit), 14th bit as 0 or 1 (bit) and 15th bit as 0 or 1(bit).
This is what I want to do:
0000 0111 1000 --> 12 bits as 120 (int)
1 <-- 13th bit
0 <-- 14th bit
1 <-- 15th bit
1 <-- 16th bit
Expected output => 0000 0111 1000 1011 : final binary and convert it to NSData.
How can I do that? Please give me some advice. Thanks all.
0000011110001011 in bit is 0x07 0x8b in byte.
unsigned char a[2] ;
a[0] = 0x07 ;
a[1] = 0x8b ;
NSData * d = [NSData dataWithBytes:a length:2] ;
Recently i wrote this code for my own project.. Check if this can be helpful to you.
//Create a NSMutableData with particular num of data bytes
NSMutableData *dataBytes= [[NSMutableData alloc] initWithLength:numberOfDataBytes];
//get the byte in which you want to change bit.
char x;
[dataBytes getBytes:&x range:NSMakeRange(bytePos,1)];
//change the bit here by shift operation
x |= 1<< (bitNum%8);
//put byte back in NSMutableData
[dataBytes replaceBytesInRange:NSMakeRange(bytePos,1) withBytes:&x length:1];
Let me know if you need more help ..:)
For changing 13th bit based on string equality
if([myString isEqualTo:#"hello"])
{
char x;
[dataBytes getBytes:&x range:NSMakeRange(0,1)];
//change the bit here by shift operation
x |= 1<< (4%8); // x |=1<<4;
//put byte back in NSMutableData
[dataBytes replaceBytesInRange:NSMakeRange(0,1) withBytes:&x length:1];
}
This is the exact code you might need:
uint16_t x= 120; //2 byte unsigned int (0000 0000 0111 1000)
//You need 13th bit to be 1
x<<=1; //Shift bits to left .(x= 0000 0000 1111 0000)
x|=1; //OR with 1 (x= 0000 0000 1111 0001)
//14th bit to be 0.
x<<=1; // (x=0000 0001 1110 0010)
//15th bit to be 1
x<<=1; //x= 0000 0011 1100 0100
x|=1; //x= 0000 0011 1100 0101
//16th bit to be 1
x<<=1; //x= 0000 0111 1000 1010
x|=1; //x= 0000 0111 1000 1011
//Now convert x into NSData
/** **** Replace this for Big Endian ***************/
NSMutableData *data = [[NSMutableData alloc] init];
int MSB = x/256;
int LSB = x%256;
[data appendBytes:&MSB length:1];
[data appendBytes:&LSB length:1];
/** **** Replace upto here.. :) ***************/
//replace with :
//NSMutableData *data = [[NSMutableData alloc] initWithBytes:&x length:sizeof(x)];
NSLog(#"%#",[data description]);
Output: <078b> //x= 0000 0111 1000 1011 //for Big Endian : <8b07> x= 1000 1011 0000 0111
I'd like to get some insight about how constant memory is allocated (using CUDA 4.2). I know that the total available constant memory is 64KB. But when is this memory actually allocated on the device? Is this limit apply to each kernel, cuda context or for the whole application?
Let's say there are several kernels in a .cu file, each using less than 64K constant memory. But the total constant memory usage is more than 64K. Is it possible to call these kernels sequentially? What happens if they are called concurrently using different streams?
What happens if there is a large CUDA dynamic library with lots of kernels each using different amounts of constant memory?
What happens if there are two applications each requiring more than half of the available constant memory? The first application runs fine, but when will the second app fail? At app start, at cudaMemcpyToSymbol() calls or at kernel execution?
Parallel Thread Execution ISA Version 3.1 section 5.1.3 discusses constant banks.
Constant memory is restricted in size, currently limited to 64KB which
can be used to hold statically-sized constant variables. There is an
additional 640KB of constant memory, organized as ten independent 64KB
regions. The driver may allocate and initialize constant buffers in
these regions and pass pointers to the buffers as kernel function
parameters. Since the ten regions are not contiguous, the driver
must ensure that constant buffers are allocated so that each buffer
fits entirely within a 64KB region and does not span a region
boundary.
A simple program can be used to illustrate the use of constant memory.
__constant__ int kd_p1;
__constant__ short kd_p2;
__constant__ char kd_p3;
__constant__ double kd_p4;
__constant__ float kd_floats[8];
__global__ void parameters(int p1, short p2, char p3, double p4, int* pp1, short* pp2, char* pp3, double* pp4)
{
*pp1 = p1;
*pp2 = p2;
*pp3 = p3;
*pp4 = p4;
return;
}
__global__ void constants(int* pp1, short* pp2, char* pp3, double* pp4)
{
*pp1 = kd_p1;
*pp2 = kd_p2;
*pp3 = kd_p3;
*pp4 = kd_p4;
return;
}
Compile this for compute_30, sm_30 and execute cuobjdump -sass <executable or obj> to disassemble you should see
Fatbin elf code:
================
arch = sm_30
code version = [1,6]
producer = cuda
host = windows
compile_size = 32bit
identifier = c:/dev/constant_banks/kernel.cu
code for sm_30
Function : _Z10parametersiscdPiPsPcPd
/*0008*/ /*0x10005de428004001*/ MOV R1, c [0x0] [0x44]; // stack pointer
/*0010*/ /*0x40001de428004005*/ MOV R0, c [0x0] [0x150]; // pp1
/*0018*/ /*0x50009de428004005*/ MOV R2, c [0x0] [0x154]; // pp2
/*0020*/ /*0x0001dde428004005*/ MOV R7, c [0x0] [0x140]; // p1
/*0028*/ /*0x13f0dc4614000005*/ LDC.U16 R3, c [0x0] [0x144]; // p2
/*0030*/ /*0x60011de428004005*/ MOV R4, c [0x0] [0x158]; // pp3
/*0038*/ /*0x70019de428004005*/ MOV R6, c [0x0] [0x15c]; // pp4
/*0048*/ /*0x20021de428004005*/ MOV R8, c [0x0] [0x148]; // p4
/*0050*/ /*0x30025de428004005*/ MOV R9, c [0x0] [0x14c]; // p4
/*0058*/ /*0x1bf15c0614000005*/ LDC.U8 R5, c [0x0] [0x146]; // p3
/*0060*/ /*0x0001dc8590000000*/ ST [R0], R7; // *pp1 = p1
/*0068*/ /*0x0020dc4590000000*/ ST.U16 [R2], R3; // *pp2 = p2
/*0070*/ /*0x00415c0590000000*/ ST.U8 [R4], R5; // *pp3 = p3
/*0078*/ /*0x00621ca590000000*/ ST.64 [R6], R8; // *pp4 = p4
/*0088*/ /*0x00001de780000000*/ EXIT;
/*0090*/ /*0xe0001de74003ffff*/ BRA 0x90;
/*0098*/ /*0x00001de440000000*/ NOP CC.T;
/*00a0*/ /*0x00001de440000000*/ NOP CC.T;
/*00a8*/ /*0x00001de440000000*/ NOP CC.T;
/*00b0*/ /*0x00001de440000000*/ NOP CC.T;
/*00b8*/ /*0x00001de440000000*/ NOP CC.T;
...........................................
Function : _Z9constantsPiPsPcPd
/*0008*/ /*0x10005de428004001*/ MOV R1, c [0x0] [0x44]; // stack pointer
/*0010*/ /*0x00001de428004005*/ MOV R0, c [0x0] [0x140]; // p1
/*0018*/ /*0x10009de428004005*/ MOV R2, c [0x0] [0x144]; // p2
/*0020*/ /*0x0001dde428004c00*/ MOV R7, c [0x3] [0x0]; // kd_p1
/*0028*/ /*0x13f0dc4614000c00*/ LDC.U16 R3, c [0x3] [0x4]; // kd_p2
/*0030*/ /*0x20011de428004005*/ MOV R4, c [0x0] [0x148]; // p3
/*0038*/ /*0x30019de428004005*/ MOV R6, c [0x0] [0x14c]; // p4
/*0048*/ /*0x20021de428004c00*/ MOV R8, c [0x3] [0x8]; // kd_p4
/*0050*/ /*0x30025de428004c00*/ MOV R9, c [0x3] [0xc]; // kd_p4
/*0058*/ /*0x1bf15c0614000c00*/ LDC.U8 R5, c [0x3] [0x6]; // kd_p3
/*0060*/ /*0x0001dc8590000000*/ ST [R0], R7;
/*0068*/ /*0x0020dc4590000000*/ ST.U16 [R2], R3;
/*0070*/ /*0x00415c0590000000*/ ST.U8 [R4], R5;
/*0078*/ /*0x00621ca590000000*/ ST.64 [R6], R8;
/*0088*/ /*0x00001de780000000*/ EXIT;
/*0090*/ /*0xe0001de74003ffff*/ BRA 0x90;
/*0098*/ /*0x00001de440000000*/ NOP CC.T;
/*00a0*/ /*0x00001de440000000*/ NOP CC.T;
/*00a8*/ /*0x00001de440000000*/ NOP CC.T;
/*00b0*/ /*0x00001de440000000*/ NOP CC.T;
/*00b8*/ /*0x00001de440000000*/ NOP CC.T;
.....................................
I annotated to the right of the SASS.
On sm30 you can see that parameters are passed in constant bank 0 starting at offset 0x140.
User defined __constant__ variables are defined in constant bank 3.
If you execute cuobjdump --dump-elf <executable or obj> you can find other interesting constant information.
32bit elf: abi=6, sm=30, flags = 0x1e011e
Sections:
Index Offset Size ES Align Type Flags Link Info Name
1 34 142 0 1 STRTAB 0 0 0 .shstrtab
2 176 19b 0 1 STRTAB 0 0 0 .strtab
3 314 d0 10 4 SYMTAB 0 2 a .symtab
4 3e4 50 0 4 CUDA_INFO 0 3 b .nv.info._Z9constantsPiPsPcPd
5 434 30 0 4 CUDA_INFO 0 3 0 .nv.info
6 464 90 0 4 CUDA_INFO 0 3 a .nv.info._Z10parametersiscdPiPsPcPd
7 4f4 160 0 4 PROGBITS 2 0 a .nv.constant0._Z10parametersiscdPiPsPcPd
8 654 150 0 4 PROGBITS 2 0 b .nv.constant0._Z9constantsPiPsPcPd
9 7a8 30 0 8 PROGBITS 2 0 0 .nv.constant3
a 7d8 c0 0 4 PROGBITS 6 3 a00000b .text._Z10parametersiscdPiPsPcPd
b 898 c0 0 4 PROGBITS 6 3 a00000c .text._Z9constantsPiPsPcPd
.section .strtab
.section .shstrtab
.section .symtab
index value size info other shndx name
0 0 0 0 0 0 (null)
1 0 0 3 0 a .text._Z10parametersiscdPiPsPcPd
2 0 0 3 0 7 .nv.constant0._Z10parametersiscdPiPsPcPd
3 0 0 3 0 b .text._Z9constantsPiPsPcPd
4 0 0 3 0 8 .nv.constant0._Z9constantsPiPsPcPd
5 0 0 3 0 9 .nv.constant3
6 0 4 1 0 9 kd_p1
7 4 2 1 0 9 kd_p2
8 6 1 1 0 9 kd_p3
9 8 8 1 0 9 kd_p4
10 16 32 1 0 9 kd_floats
11 0 192 12 10 a _Z10parametersiscdPiPsPcPd
12 0 192 12 10 b _Z9constantsPiPsPcPd
The kernel parameter constant bank is versioned per launch so that concurrent kernels can be executed. The compiler and user constants are per CUmodule. It is the responsibility of the developer to manage coherency of this data. For example, the developer has to ensure that a cudaMemcpyToSymbol is update in a safe manner.