Difference between revisions of "GPU/Shader Instruction Set"

From 3dbrew
< GPU
Jump to navigation Jump to search
(FLR !)
(Clarify LITP instruction)
 
(79 intermediate revisions by 11 users not shown)
Line 1: Line 1:
[[Category:GFX]]
+
[[Category:GPU]]
  
 
== Overview ==
 
== Overview ==
A compiled shader binary is comprised of two parts : the main instruction sequence and the operand descriptor table. These are both sent to the GPU around the same time but using separate [[GPU Commands]]. Instructions (such as format 1 instruction) may reference operand descriptors. When such is the case, the operand descriptor ID is the offset, in words, of the descriptor within the table.
+
A compiled shader binary is comprised of two parts : the main instruction sequence and the operand descriptor table. These are both sent to the GPU around the same time but using separate [[GPU/Internal_Registers|GPU Commands]]. Instructions (such as format 1 instruction) may reference operand descriptors. When such is the case, the operand descriptor ID is the offset, in words, of the descriptor within the table.
 
Both instructions and descriptors are coded in little endian.
 
Both instructions and descriptors are coded in little endian.
Basic implementations of the following specification can be found at [https://github.com/smealum/aemstro] and [https://github.com/neobrain/nihstro]
+
Basic implementations of the following specification can be found at [https://github.com/smealum/aemstro] and [https://github.com/neobrain/nihstro].
 +
The instruction set seems to have been heavily inspired by Microsoft's vs_3_0 [http://msdn.microsoft.com/en-us/library/windows/desktop/bb172938%28v=vs.85%29.aspx] and the Direct3D shader code [https://msdn.microsoft.com/en-us/library/windows/hardware/ff552891%28v=vs.85%29.aspx].
 
Please note that this page is being written as the instruction set is reverse engineered; as such it may very well contain mistakes.
 
Please note that this page is being written as the instruction set is reverse engineered; as such it may very well contain mistakes.
 +
 +
Debug information found in the code.bin of "Ironfall: Invasion" suggests that there may not be more than 512 instructions and 128 operand descriptors in a shader.
 +
 +
== Nomenclature ==
 +
 +
* opcode names with I appended to them are the same as their non-I version, except they use the inverted instruction format, giving 7 bits to SRC2 (and access to constant registers) and 5 bits to SRC1
 +
 +
* opcode names with U appended to them are the same as their non-U version, except they are executed conditionally based on the value of a constant boolean register.
 +
 +
* opcode names with C appended to them are the same as their non-C version, except they are executed conditionally based on a logical expression specified in the instruction.
  
 
== Instruction formats ==
 
== Instruction formats ==
Line 30: Line 41:
 
|  0x13
 
|  0x13
 
|  0x2
 
|  0x2
|  Address register index (IDX)
+
|  Address register index for SRC1 (IDX_1)
 
|-
 
|-
 
|  0x15
 
|  0x15
Line 54: Line 65:
 
|  0x7
 
|  0x7
 
|  0x7
 
|  0x7
|  Source 1 register (SRC1)
+
|  Source 2 register (SRC2)
 
|-
 
|-
 
|  0xE
 
|  0xE
 
|  0x5
 
|  0x5
|  Source 2 register (SRC2)
+
|  Source 1 register (SRC1)
 
|-
 
|-
 
|  0x13
 
|  0x13
 
|  0x2
 
|  0x2
|  Address register index (IDX)
+
|  Address register index for SRC2 (IDX_2)
 
|-
 
|-
 
|  0x15
 
|  0x15
Line 90: Line 101:
 
|  0x13
 
|  0x13
 
|  0x2
 
|  0x2
| Address register index (IDX)
+
|   Address register index for SRC1 (IDX_1)
 
|-
 
|-
 
|  0x15
 
|  0x15
Line 113: Line 124:
 
|-
 
|-
 
|  0x7
 
|  0x7
 +
|  0x5
 +
|  Source 2 register (SRC2)
 +
|-
 +
|  0xC
 
|  0x7
 
|  0x7
 
|  Source 1 register (SRC1)
 
|  Source 1 register (SRC1)
|-
 
|  0xE
 
|  0x5
 
|  Source 2 register (SRC2)
 
 
|-
 
|-
 
|  0x13
 
|  0x13
 
|  0x2
 
|  0x2
|  Address register index (IDX)
+
|  Address register index for SRC1 (IDX_1)
 
|-
 
|-
 
|  0x15
 
|  0x15
Line 158: Line 169:
 
|  0x18
 
|  0x18
 
|  0x1
 
|  0x1
|  Y negation bit (NEGY)
+
|  Y reference bit (REFY)
 
|-
 
|-
 
|  0x19
 
|  0x19
 
|  0x1
 
|  0x1
|  X negation bit (NEGX)
+
|  X reference bit (REFX)
 
|-
 
|-
 
|  0x1A
 
|  0x1A
Line 169: Line 180:
 
|}
 
|}
  
Format 3 : (used for uniform-based conditional flow control instructions)
+
Format 3 : (used for constant-based conditional flow control instructions)
 
{| class="wikitable" border="1"
 
{| class="wikitable" border="1"
 
|-
 
|-
Line 186: Line 197:
 
|  0x16
 
|  0x16
 
|  0x4
 
|  0x4
Uniform ID (BOOL/INT)
+
Constant ID (BOOL/INT)
 
|-
 
|-
 
|  0x1A
 
|  0x1A
Line 201: Line 212:
 
|-
 
|-
 
|  0x16
 
|  0x16
0x2
+
0x1
|  Primitive ID (PRIMID)
+
|  Winding flag (FLAG_WINDING)
 +
|-
 +
|  0x17
 +
|  0x1
 +
|  Primitive emit flag (FLAG_PRIMEMIT)
 
|-
 
|-
 
|  0x18
 
|  0x18
Line 233: Line 248:
 
|-
 
|-
 
|  0x11
 
|  0x11
 +
|  0x5
 +
|  Source 1 register (SRC1)
 +
|-
 +
|  0x16
 +
|  0x2
 +
|  Address register index for SRC2 (IDX_2)
 +
|-
 +
|  0x18
 +
|  0x5
 +
|  Destination register (DST)
 +
|-
 +
|  0x1D
 +
|  0x3
 +
|  Opcode
 +
|}
 +
 +
Format 5i : (used for MADI)
 +
{| class="wikitable" border="1"
 +
|-
 +
!  Offset
 +
!  Size (bits)
 +
!  Description
 +
|-
 +
|  0x0
 +
|  0x5
 +
|  Operand descriptor ID (DESC)
 +
|-
 +
|  0x5
 
|  0x7
 
|  0x7
 +
|  Source 3 register (SRC3)
 +
|-
 +
|  0xC
 +
|  0x5
 +
|  Source 2 register (SRC2)
 +
|-
 +
|  0x11
 +
|  0x5
 
|  Source 1 register (SRC1)
 
|  Source 1 register (SRC1)
 +
|-
 +
|  0x16
 +
|  0x2
 +
|  Address register index for SRC3 (IDX_3)
 
|-
 
|-
 
|  0x18
 
|  0x18
Line 246: Line 301:
  
 
== Instructions ==
 
== Instructions ==
 +
Unless noted otherwise, SRC1 and SRC2 refer to their respectively indexed float[4] registers (after swizzling). Similarly, DST refers to its indexed register modulo destination component masking, i.e. an expression like DST=SRC1 might actually just set DST.y to SRC1.y.
 +
  
 
{| class="wikitable" border="1"
 
{| class="wikitable" border="1"
Line 257: Line 314:
 
|  1
 
|  1
 
|  ADD
 
|  ADD
|  Adds two vectors component by component; DST[i] = SRC1[i]+SRC2[i] for all i (modulo destination component masking)
+
|  Adds two vectors component by component; DST[i] = SRC1[i]+SRC2[i] for all i
 
|-
 
|-
 
|  0x01
 
|  0x01
Line 271: Line 328:
 
|  0x03
 
|  0x03
 
|  1
 
|  1
???
+
DPH
?
+
Computes dot product on a 3-component vector with 1.0 appended to it and a 4-component vector; DST = SRC1.SRC2 (with SRC1 homogenous)
 +
|-
 +
|  0x04
 +
|  1
 +
|  DST
 +
|  Equivalent to Microsoft's [https://msdn.microsoft.com/en-us/library/windows/desktop/bb219790.aspx dst] instruction: DST = {1, SRC1[1]*SRC2[1], SRC1[2], SRC2[3]}
 
|-
 
|-
 
|  0x05
 
|  0x05
1
+
1u
???
+
EX2
?
+
Computes SRC1's first component exponent with base 2; DST[i] = EXP2(SRC1[0]) for all i
 
|-
 
|-
 
|  0x06
 
|  0x06
1
+
1u
???
+
LG2
?
+
Computes SRC1's first component logarithm with base 2; DST[i] = LOG2(SRC1[0]) for all i
 +
|-
 +
|  0x07
 +
|  1u
 +
|  LITP
 +
|  Partial lighting computation, may be used in conjunction with EX2, LG2, etc to compute the vertex lighting coefficients. See the [https://msdn.microsoft.com/en-us/library/windows/desktop/bb174703.aspx Microsoft] and [https://registry.khronos.org/OpenGL/extensions/ARB/ARB_vertex_program.txt ARB] docs for more information on how to implement the full lit function; DST = {max(src.x, 0), max(min(src.y, 127.9961), -127.9961), 0, max(src.w, 0)} and it sets the cmp.x and cmp.y flags based on if the respective src.x and src.w components are >= 0.
 
|-
 
|-
 
|  0x08
 
|  0x08
 
|  1
 
|  1
 
|  MUL
 
|  MUL
|  Multiplies two vectors component by component; DST[i] = SRC1[i].SRC2[i] for all i (modulo destination component masking)
+
|  Multiplies two vectors component by component; DST[i] = SRC1[i].SRC2[i] for all i
 
|-
 
|-
 
|  0x09
 
|  0x09
 
|  1
 
|  1
???
+
SGE
|  ?
+
Sets output if SRC1 is greater than or equal to SRC2; DST[i] = (SRC1[i] >= SRC2[i]) ? 1.0 : 0.0 for all i
 
|-
 
|-
 
|  0x0A
 
|  0x0A
 
|  1
 
|  1
???
+
SLT
|  ?
+
Sets output if SRC1 is strictly less than SRC2; DST[i] = (SRC1[i] < SRC2[i]) ? 1.0 : 0.0 for all i
 
|-
 
|-
 
|  0x0B
 
|  0x0B
 
|  1u
 
|  1u
 
|  FLR
 
|  FLR
|  Computes SRC1's floor component by component; DST[i] = FLOOR(SRC1[i]) for all i (modulo destination component masking)
+
|  Computes SRC1's floor component by component; DST[i] = FLOOR(SRC1[i]) for all i
 
|-
 
|-
 
|  0x0C
 
|  0x0C
 
|  1
 
|  1
 
|  MAX
 
|  MAX
|  Takes the max of two vectors, component by component; DST[i] = MAX(SRC1[i], SRC2[i]) for all i (modulo destination component masking)
+
|  Takes the max of two vectors, component by component; DST[i] = MAX(SRC1[i], SRC2[i]) for all i
 
|-
 
|-
 
|  0x0D
 
|  0x0D
 
|  1
 
|  1
 
|  MIN
 
|  MIN
|  Takes the min of two vectors, component by component; DST[i] = MIN(SRC1[i], SRC2[i]) for all i (modulo destination component masking)
+
|  Takes the min of two vectors, component by component; DST[i] = MIN(SRC1[i], SRC2[i]) for all i
 
|-
 
|-
 
|  0x0E
 
|  0x0E
1
+
1u
 
|  RCP
 
|  RCP
|  Computes the reciprocal of the vector, component by component; DST[i] = 1/SRC1[i] for all i (modulo destination component masking)
+
|  Computes the reciprocal of the vector's first component; DST[i] = 1/SRC1[0] for all i
 
|-
 
|-
 
|  0x0F
 
|  0x0F
1
+
1u
 
|  RSQ
 
|  RSQ
|  Computes the reciprocal of the square root of the vector, component by component; DST[i] = 1/sqrt(SRC1[i]) for all i (modulo destination component masking)
+
|  Computes the reciprocal of the square root of the vector's first component; DST[i] = 1/sqrt(SRC1[0]) for all i
 +
|-
 +
| 0x10
 +
| ?
 +
| ???
 +
| ?
 +
|-
 +
| 0x11
 +
| ?
 +
| ???
 +
| ?
 
|-
 
|-
 
|  0x12
 
|  0x12
1
+
1u
ARL
+
MOVA
Address Register Load; sets (a0, a1, _, _) to SRC1 (cast to integer).
+
Move to address register; Casts the float value given by SRC1 to an integer (truncating the fractional part) and assigns the result to (a0.x, a0.y, _, _), respecting the destination component mask.
 
|-
 
|-
 
|  0x13
 
|  0x13
Line 333: Line 410:
 
|  MOV
 
|  MOV
 
|  Moves value from one register to another; DST = SRC1.
 
|  Moves value from one register to another; DST = SRC1.
 +
|-
 +
|  0x14
 +
|  ?
 +
|  ???
 +
|  ?
 +
|-
 +
|  0x15
 +
|  ?
 +
|  ???
 +
|  ?
 +
|-
 +
|  0x16
 +
|  ?
 +
|  ???
 +
|  ?
 +
|-
 +
|  0x17
 +
|  ?
 +
|  ???
 +
|  ?
 
|-
 
|-
 
|  0x18
 
|  0x18
 
|  1i
 
|  1i
DP4I?
+
DPHI
|  Computes dot product on 4-component vectors; DST = SRC1.SRC2 ?
+
|  Computes dot product on a 3-component vector with 1.0 appended to it and a 4-component vector; DST = SRC1.SRC2 (with SRC1 homogenous)
 +
|-
 +
|  0x19
 +
|  1i
 +
|  DSTI
 +
|  DST with sources swapped.
 
|-
 
|-
 
|  0x1A
 
|  0x1A
|  1
+
1i
 +
|  SGEI
 +
|  Sets output if SRC1 is greater than or equal to SRC2; DST[i] = (SRC1[i] >= SRC2[i]) ? 1.0 : 0.0 for all i
 +
|-
 +
|  0x1B
 +
|  1i
 +
|  SLTI
 +
|  Sets output if SRC1 is strictly less than SRC2; DST[i] = (SRC1[i] < SRC2[i]) ? 1.0 : 0.0 for all i
 +
|-
 +
|  0x1C
 +
|  ?
 +
|  ???
 +
|  ?
 +
|-
 +
|  0x1D
 +
|  ?
 +
|  ???
 +
|  ?
 +
|-
 +
|  0x1E
 +
|  ?
 
|  ???
 
|  ???
 
|  ?
 
|  ?
 
|-
 
|-
0x1B
+
0x1F
1
+
?
 
|  ???
 
|  ???
 
|  ?
 
|  ?
 +
|-
 +
|  0x20
 +
|  0
 +
|  BREAK
 +
|  Breaks out of LOOP block; do not use while in nested IF/CALL block inside LOOP block.
 
|-
 
|-
 
|  0x21
 
|  0x21
1
+
0
END2
+
NOP
?
+
Does literally nothing.
 
|-
 
|-
 
|  0x22
 
|  0x22
1
+
0
END1
+
END
?
+
Signals the shader unit that processing for this vertex/primitive is done.
 
|-
 
|-
 
|  0x23
 
|  0x23
Line 391: Line 518:
 
|  0x29
 
|  0x29
 
|  3
 
|  3
FORLOOP
+
LOOP
|  Loops over the code between itself and DST. Increments lcnt after each loop. Stops looping once lcnt reaches the value contained by the integer uniform specified by INT. (i0-i7)
+
|  Loops over the code between itself and DST (inclusive), performing INT.x+1 iterations in total. First, aL is initialized to INT.y. After each iteration, aL is incremented by INT.z.
 
|-
 
|-
 
|  0x2A
 
|  0x2A
 
|  0 (no param)
 
|  0 (no param)
 
|  EMIT
 
|  EMIT
|  (geometry shader only) Emits a vertex (and primitive if PRIMID is non-zero). SETEMIT must be called before this.
+
|  (geometry shader only) Emits a vertex (and primitive if FLAG_PRIMEMIT was set in the corresponding SETEMIT). SETEMIT must be called before this.
 
|-
 
|-
 
|  0x2B
 
|  0x2B
 
|  4
 
|  4
 
|  SETEMIT
 
|  SETEMIT
|  (geometry shader only) Sets VTXID and PRIMID for the next EMIT instruction. VTXID is the ID of the vertex about to be emitted within the primitive, while PRIMID is zero if we are just emitting a single vertex and non-zero if are emitting a vertex and primitive simultaneously. Note that the output vertex buffer (which holds 4 vertices) is '''not''' cleared when the primitive is emitted, meaning that vertices from the previous primitive can be reused for the current one. (this is still a working hypothesis and unconfirmed)
+
|  (geometry shader only) Sets VTXID, FLAG_WINDING and FLAG_PRIMEMIT for the next EMIT instruction. VTXID is the ID of the vertex about to be emitted within the primitive, while FLAG_PRIMEMIT is zero if we are just emitting a single vertex and non-zero if are emitting a vertex and primitive simultaneously. FLAG_WINDING controls the output primitive's winding. Note that the output vertex buffer (which holds 4 vertices) is '''not''' cleared when the primitive is emitted, meaning that vertices from the previous primitive can be reused for the current one. (this is still a working hypothesis and unconfirmed)
 
|-
 
|-
 
|  0x2C
 
|  0x2C
Line 412: Line 539:
 
|  3
 
|  3
 
|  JMPU
 
|  JMPU
|  If condition BOOL is true, then jumps to DST, else does nothing. It seems possible that having NUM = 1 will jump if BOOL is false instead, though this is unconfirmed.
+
|  If condition BOOL is true, then jumps to DST, else does nothing. Having bit 0 of NUM = 1 will invert the test, jumping if BOOL is false instead.
 
|-
 
|-
 
|  0x2E-0x2F
 
|  0x2E-0x2F
 
|  1c
 
|  1c
 
|  CMP
 
|  CMP
|  Sets booleans cmp.x and cmp.y based on the operand's x and y components and the CMPX and CMPY comparison operators respectively. See [[#Comparison_operator|below]] for details about operators.
+
|  Sets booleans cmp.x and cmp.y based on the operand's x and y components and the CMPX and CMPY comparison operators respectively. See [[#Comparison_operator|below]] for details about operators. It's unknown whether CMP respects the destination component mask or not.
 
|-
 
|-
 
|  0x30-0x37
 
|  0x30-0x37
5
+
5i
LRP
+
MADI
Does linear interpolation between two vectors, using a third as the interpolation factor, component by component; DST[i] = SRC1[i].(1.0 - SRC3[i]) + SRC2[i].(SRC3[i]) for all i (modulo destination component masking)
+
Multiplies two vectors and adds a third one component by component; DST[i] = SRC3[i] + SRC2[i].SRC1[i] for all i; this is not an FMA, the intermediate result is rounded
 
|-
 
|-
 
|  0x38-0x3F
 
|  0x38-0x3F
 
|  5
 
|  5
 
|  MAD
 
|  MAD
|  Multiplies two vectors and adds a third one component by component; DST[i] = SRC3[i] + SRC2[i].SRC1[i] for all i (modulo destination component masking)
+
|  Multiplies two vectors and adds a third one component by component; DST[i] = SRC3[i] + SRC2[i].SRC1[i] for all i; this is not an FMA, the intermediate result is rounded
 
|}
 
|}
  
Line 511: Line 638:
  
 
The component selector enables swizzling. For example, component selector 0x1B is equivalent to .xyzw, while 0x55 is equivalent to .yyyy.
 
The component selector enables swizzling. For example, component selector 0x1B is equivalent to .xyzw, while 0x55 is equivalent to .yyyy.
 +
 +
Depending on the current shader opcode, source components are disabled implicitly by setting the destination component mask. For example, ADD o0.xy, r0.xyzw, r1.xyzw will not make use of r0's or r1's z/w components, while DP4 o0.xy, r0.xyzw, r1.xyzw will use all input components regardless of the used destination component mask.
  
 
== Relative addressing ==
 
== Relative addressing ==
  
There are 3 global address registers : a0, a1 and a2 = lcnt (loop counter). For format 1 instructions, when IDX != 0, the value of the corresponding address register is added to SRC1's value.
+
{| class="wikitable" border="1"
 +
|-
 +
IDX raw value
 +
! Register name
 +
|-
 +
|  0x0
 +
|  None
 +
|-
 +
|  0x1
 +
|  a0.x
 +
|-
 +
|  0x2
 +
|  a0.y
 +
|-
 +
|  0x3
 +
|  aL
 +
|}
  
For example, if IDX = 2, a1 = 3 and SRC1 = c8, then instead SRC1+a1 = c11 will be used for the instruction.
+
There are 3 address registers: a0.x, a0.y and aL (loop counter). For format 1 instructions, when IDX != 0, the value of the corresponding address register is added to SRC1's value. For example, if IDX = 2, a0.y = 3 and SRC1 = c8, then instead SRC1+a0.y = c11 will be used for the instruction. It is only possible to use address registers on constant registers, attempting to use them on input attribute or temporary registers results in the address register being ignored (i.e. read as zero).
  
a0 and a1 can be set manually through the ARL instruction. lcnt is set automatically by the LOOP instruction. Note that lcnt is still accessible and valid after exiting a LOOP block.
+
a0.x and a0.y are set manually through the MOVA instruction by rounding a float value to integer precision. Hence, they may take negative values. The way out-of-bounds values behave when reading uniforms is as follows:
 +
* If the offset is out of byte bounds (less than -128 or greater than 127), the offset is not applied (treated as 0).
 +
* The offset is added to the constant register index and masked by 0x7F.
 +
* If the resulting index is greater than 95, the result is (1, 1, 1, 1).
 +
* Otherwise, the result is the value at the indexed constant register.
 +
 
 +
aL can only be set indirectly by the LOOP instruction. It is still accessible and valid after exiting a LOOP block, though.
  
 
== Comparison operator ==
 
== Comparison operator ==
Line 567: Line 718:
 
A number of format 2 instructions are executed conditionally. These conditions are based on two boolean registers which can be set with CMP : cmp.x and cmp.y.
 
A number of format 2 instructions are executed conditionally. These conditions are based on two boolean registers which can be set with CMP : cmp.x and cmp.y.
  
Conditional instructions include 3 parameters : CONDOP, NEGX and NEGY. NEGX and NEGY determine whether the conditional expression will use cmp.x or !cmp.x, and cmp.y or !cmp.y respectively (NEGX set means we use cmp.x, and NEGX not set means we use !cmp.x). CONDOP describes the actual expression. There are four conditional expression formats :
+
Conditional instructions include 3 parameters : CONDOP, REFX and REFY. REFX and REFY are reference values which are tested for equality against cmp.x and cmp.y, respectively. CONDOP describes how the final truth value is constructed from the results of the two tests. There are four conditional expression formats :
  
 
{| class="wikitable" border="1"
 
{| class="wikitable" border="1"
Line 576: Line 727:
 
|-
 
|-
 
|  0x0
 
|  0x0
|  <nowiki>[!]cmp.x || [!]cmp.y</nowiki>
+
|  <nowiki>cmp.x == REFX || cmp.y == REFY</nowiki>
 
|  OR
 
|  OR
 
|-
 
|-
 
|  0x1
 
|  0x1
|  <nowiki>[!]cmp.x && [!]cmp.y</nowiki>
+
|  <nowiki>cmp.x == REFX && cmp.y == REFY</nowiki>
 
|  AND
 
|  AND
 
|-
 
|-
 
|  0x2
 
|  0x2
[!]cmp.X
+
|  cmp.x == REFX
 
|  X
 
|  X
 
|-
 
|-
 
|  0x3
 
|  0x3
[!]cmp.y
+
|  cmp.y == REFY
 
|  Y
 
|  Y
 
|}
 
|}
 
For example, with CONDOP=1, NEGX=1 and NEGY=0, the resulting expression would be (cmp.x && !cmp.y).
 
  
 
== Registers ==
 
== Registers ==
Most registers (all the ones within the 0x00-0x7F range) are float[4] vectors. There are also boolean registers (b0-b7) and integer registers (i0-i7). How the latter ones are set is as of yet unknown.
 
  
Attribute (input, RO) registers are located within the 0x0-0xF range. What data they are fed is specified by the CPU.
+
{| class="wikitable" border="1"
 
+
|-
Output (WO) registers are also located within the 0x0-0xF range. What type of data they are contain is specified by the CPU.
+
!  Name
 +
!  Format
 +
!  Type
 +
!  Access
 +
!  Written by
 +
!  Description
 +
|-
 +
|  v0-v15
 +
|  vector
 +
|  float
 +
|  Read only
 +
|  Application/Vertex-stream
 +
|  Input registers.
 +
|-
 +
|  o0-o15
 +
|  vector
 +
|  float
 +
|  Write only
 +
|  Vertex shader
 +
|  Output registers.
 +
|-
 +
|  r0-r15
 +
|  vector
 +
|  float
 +
|  Read/Write
 +
|  Vertex shader
 +
|  Temporary registers.
 +
|-
 +
|  c0-c95
 +
|  vector
 +
|  float
 +
|  Read only
 +
|  Application/Vertex-stream
 +
|  Floating-point Constant registers.
 +
|-
 +
|  i0-i3
 +
|  vector
 +
|  integer
 +
|  Read only
 +
|  Application
 +
|  Integer Constant registers. (special purpose)
 +
|-
 +
|  b0-b15
 +
|  scalar
 +
|  boolean
 +
|  Read only
 +
|  Application
 +
|  Boolean Constant registers. (special purpose)
 +
|-
 +
|  a0.x & a0.y
 +
|  scalar
 +
|  integer
 +
|  Use/Write
 +
|  Vertex shader
 +
|  Address registers.
 +
|-
 +
|  aL
 +
|  scalar
 +
|  integer
 +
|  Use
 +
|  Vertex shader
 +
|  Loop count register.
 +
|}
  
Temporary (RW) register are located within the 0x10-0x1F range. They can contain any type of data.
+
Input attribute registers store the per-vertex data given by the CPU and hence are read-only.
  
Uniform (RO) registers are located within the 0x20-0x7F range. Their content is set by the CPU.
+
Output registers hold the data to be passed to the later GPU stages and are write-only. Each of the output register is assigned a semantic by setting the corresponding [[GPU_Internal_Registers]]. Output registers o7-o15 are only available in vertex shaders.
 +
Keep in mind that writing to the same output register/component more than once appears appears to cause problems (e.g. GPU hangs).
  
SRC2 being only 5 bits long rather than 7 bits like its friend SRC1, it can only access v (input attribute) and r (temporary) registers.
+
Temporary registers can be used for intermediate calculations and can be both read and written.
  
Registers in the 0x88-0x97 range are uniform booleans.
+
Constant registers hold data uploaded by the application which remain constant throughout all processed vertices. There are 96 float[4] constant registers (c0-c95), eight boolean constant registers (b0-b7), and four int[4] constant registers (i0-i3).
 +
Many shader instructions which take float arguments can only provide the full 7 bits for one SRC operand. All other source operands can only be used to refer to input attributes or temporary registers and cannot be passed Floating-point Constant registers.
  
It appears that writing twice to the same output register can cause problems, such as the GPU hanging.
+
Address registers and the Loop count register can be used to to provide relative addressing for the designated SRC operand. For more information, see the section on [[#Relative_addressing|relative addressing]].
  
 
DST mapping :
 
DST mapping :
Line 619: Line 831:
 
!  Description
 
!  Description
 
|-
 
|-
|  0x0-0x7
+
|  0x0-0xF
|  o0-o7
+
|  o0-o15
 
|  Output registers.
 
|  Output registers.
 
|-
 
|-
Line 632: Line 844:
 
{| class="wikitable" border="1"
 
{| class="wikitable" border="1"
 
|-
 
|-
SRC1 raw value
+
SRC raw value
 
!  Register name
 
!  Register name
 
!  Description
 
!  Description
 
|-
 
|-
|  0x0-0x7
+
|  0x0-0xF
|  v0-v7
+
|  v0-v15
 
|  Input attribute registers.
 
|  Input attribute registers.
 
|-
 
|-
Line 646: Line 858:
 
|  0x20-0x7F
 
|  0x20-0x7F
 
|  c0-c95
 
|  c0-c95
Vector uniform registers.
+
Constant registers.
 
|}
 
|}
  
Note that 5bit SRC registers (SRC2 in format 1 for example) can't access c0-c95 because they don't have enough bits.
+
== Floating-Point Behavior ==
 +
 
 +
The PICA200 is not IEEE-compliant. It has positive and negative infinities and NaN, but does not seem to have negative 0. Input and output subnormals are flushed to +0. The internal floating point format seems to be the same as used in shader binaries: 1 sign bit, 7 exponent bits, 16 (explicit) mantissa bits. Several instructions also have behavior that differs from the IEEE functions. Here are the results from some tests done on hardware (s = largest subnormal, n = smallest positive normal):
 +
 
 +
{| class="wikitable" border="1"
 +
|-
 +
!  Computation
 +
!  Result
 +
!  Notes
 +
|-
 +
|  inf * 0
 +
|  0
 +
|  Including inside MUL, MAD, DP4, etc.
 +
|-
 +
|  NaN * 0
 +
|  NaN
 +
 +
|-
 +
|  +inf - +inf
 +
|  NaN
 +
|  Indicates +inf is real inf, not FLT_MAX
 +
|-
 +
|  rsq(rcp(-inf))
 +
|  +inf
 +
|  Indicates that there isn't -0.0.
 +
 
 +
|- style="border-top: double"
 +
|  rcp(-0)
 +
|  +inf
 +
|  no -0 so differs from IEEE where rcp(-0) = -inf
 +
|-
 +
|  rcp(0)
 +
|  +inf
 +
 +
|-
 +
|  rcp(+inf)
 +
|  0
 +
 +
|-
 +
|  rcp(NaN)
 +
|  NaN
 +
 +
 
 +
|- style="border-top: double"
 +
|  rsq(-0)
 +
|  +inf
 +
|  no -0 so differs from IEEE where rsq(-0) = -inf
 +
|-
 +
|  rsq(-2)
 +
|  NaN
 +
 +
|-
 +
|  rsq(+inf)
 +
|  0
 +
 +
|-
 +
|  rsq(-inf)
 +
|  NaN
 +
 +
|-
 +
|  rsq(NaN)
 +
|  NaN
 +
 +
 
 +
|- style="border-top: double"
 +
|  max(0, +inf)
 +
|  +inf
 +
 +
|-
 +
|  max(0, -inf)
 +
|  -inf
 +
 +
|-
 +
|  max(0, NaN)
 +
|  NaN
 +
|  max violates IEEE but match GLSL spec
 +
|-
 +
|  max(NaN, 0)
 +
|  0
 +
 +
|-
 +
|  max(-inf, +inf)
 +
|  +inf
 +
 +
 
 +
|- style="border-top: double"
 +
|  min(0, +inf)
 +
|  0
 +
 +
|-
 +
|  min(0, -inf)
 +
|  -inf
 +
 +
|-
 +
|  min(0, NaN)
 +
|  NaN
 +
|  min violates IEEE but match GLSL spec
 +
|-
 +
|  min(NaN, 0)
 +
|  0
 +
 +
|-
 +
|  min(-inf, +inf)
 +
|  -inf
 +
|
 +
 
 +
|- style="border-top: double"
 +
|  cmp(s, 0)
 +
|  false
 +
|  cmp does not flush input subnormals
 +
|-
 +
|  max(s, 0)
 +
|  s
 +
|  max does not flush input or output subnormals
 +
|-
 +
|  mul(s, 2)
 +
|  0
 +
|  input subnormals are flushed in arithmetic instructions
 +
|-
 +
|  mul(n, 0.5)
 +
|  0
 +
|  output subnormals are flushed in arithmetic instructions
 +
|}
 +
 
 +
1.0 can be multiplied 63 times by 0.5 until the result compares equal zero. This is consistent with a 7-bit exponent and output subnormal flushing.
 +
 
 +
== Control Flow ==
 +
 
 +
Control flow is implemented using four independent stacks:
 +
 
 +
* 4-deep CALL stack
 +
* 8-deep IF stack
 +
* 4-deep LOOP stack
 +
 
 +
All stacks are initially empty. After every instruction but before JMP takes effect, the PC is incremented and a copy is sent to each stack. Each stack is checked against its copy of the PC. If an entry is popped from the stack, the copied PC is updated and used for the next check of this stack, although the IF/LOOP stacks can each only pop one entry per instruction, whereas the CALL stack is checked again until it doesn't match or the stack is empty. The updated PC copy with the highest priority wins: LOOP (highest), IF, CALL, JMP, original PC (lowest).
 +
 
 +
Special cases:
 +
* JMP overwrites the PC *after* the stacks checks (and only if no stack was popped).
 +
* Executing a BREAK on an empty LOOP stack hangs the GPU.
 +
* A stack overflow discards the oldest element, so you could think of it as a queue or a ring buffer.
 +
* If the CALL stack is popped four times in a row, the fourth update to its copy of the PC is missed (the third PC update will be propagated). Probably a hardware bug.

Latest revision as of 10:38, 2 October 2023


Overview[edit]

A compiled shader binary is comprised of two parts : the main instruction sequence and the operand descriptor table. These are both sent to the GPU around the same time but using separate GPU Commands. Instructions (such as format 1 instruction) may reference operand descriptors. When such is the case, the operand descriptor ID is the offset, in words, of the descriptor within the table. Both instructions and descriptors are coded in little endian. Basic implementations of the following specification can be found at [1] and [2]. The instruction set seems to have been heavily inspired by Microsoft's vs_3_0 [3] and the Direct3D shader code [4]. Please note that this page is being written as the instruction set is reverse engineered; as such it may very well contain mistakes.

Debug information found in the code.bin of "Ironfall: Invasion" suggests that there may not be more than 512 instructions and 128 operand descriptors in a shader.

Nomenclature[edit]

  • opcode names with I appended to them are the same as their non-I version, except they use the inverted instruction format, giving 7 bits to SRC2 (and access to constant registers) and 5 bits to SRC1
  • opcode names with U appended to them are the same as their non-U version, except they are executed conditionally based on the value of a constant boolean register.
  • opcode names with C appended to them are the same as their non-C version, except they are executed conditionally based on a logical expression specified in the instruction.

Instruction formats[edit]

Format 1 : (used for register operations)

Offset Size (bits) Description
0x0 0x7 Operand descriptor ID (DESC)
0x7 0x5 Source 2 register (SRC2)
0xC 0x7 Source 1 register (SRC1)
0x13 0x2 Address register index for SRC1 (IDX_1)
0x15 0x5 Destination register (DST)
0x1A 0x6 Opcode

Format 1i : (used for register operations)

Offset Size (bits) Description
0x0 0x7 Operand descriptor ID (DESC)
0x7 0x7 Source 2 register (SRC2)
0xE 0x5 Source 1 register (SRC1)
0x13 0x2 Address register index for SRC2 (IDX_2)
0x15 0x5 Destination register (DST)
0x1A 0x6 Opcode

Format 1u : (used for unary register operations)

Offset Size (bits) Description
0x0 0x7 Operand descriptor ID (DESC)
0xC 0x7 Source 1 register (SRC1)
0x13 0x2 Address register index for SRC1 (IDX_1)
0x15 0x5 Destination register (DST)
0x1A 0x6 Opcode

Format 1c : (used for comparison operations)

Offset Size (bits) Description
0x0 0x7 Operand descriptor ID (DESC)
0x7 0x5 Source 2 register (SRC2)
0xC 0x7 Source 1 register (SRC1)
0x13 0x2 Address register index for SRC1 (IDX_1)
0x15 0x3 Comparison operator for Y (CMPY)
0x18 0x3 Comparison operator for X (CMPX)
0x1B 0x5 Opcode

Format 2 : (used for flow control instructions)

Offset Size (bits) Description
0x0 0x8 Number of instructions (NUM)
0xA 0xC Destination offset (in words) (DST)
0x16 0x2 Condition boolean operator (CONDOP)
0x18 0x1 Y reference bit (REFY)
0x19 0x1 X reference bit (REFX)
0x1A 0x6 Opcode

Format 3 : (used for constant-based conditional flow control instructions)

Offset Size (bits) Description
0x0 0x8 Number of instructions ? (NUM)
0xA 0xC Destination offset (in words) (DST)
0x16 0x4 Constant ID (BOOL/INT)
0x1A 0x6 Opcode

Format 4 : (used for SETEMIT)

Offset Size (bits) Description
0x16 0x1 Winding flag (FLAG_WINDING)
0x17 0x1 Primitive emit flag (FLAG_PRIMEMIT)
0x18 0x2 Vertex ID (VTXID)
0x1A 0x6 Opcode

Format 5 : (used for MAD)

Offset Size (bits) Description
0x0 0x5 Operand descriptor ID (DESC)
0x5 0x5 Source 3 register (SRC3)
0xA 0x7 Source 2 register (SRC2)
0x11 0x5 Source 1 register (SRC1)
0x16 0x2 Address register index for SRC2 (IDX_2)
0x18 0x5 Destination register (DST)
0x1D 0x3 Opcode

Format 5i : (used for MADI)

Offset Size (bits) Description
0x0 0x5 Operand descriptor ID (DESC)
0x5 0x7 Source 3 register (SRC3)
0xC 0x5 Source 2 register (SRC2)
0x11 0x5 Source 1 register (SRC1)
0x16 0x2 Address register index for SRC3 (IDX_3)
0x18 0x5 Destination register (DST)
0x1D 0x3 Opcode

Instructions[edit]

Unless noted otherwise, SRC1 and SRC2 refer to their respectively indexed float[4] registers (after swizzling). Similarly, DST refers to its indexed register modulo destination component masking, i.e. an expression like DST=SRC1 might actually just set DST.y to SRC1.y.


Opcode Format Name Description
0x00 1 ADD Adds two vectors component by component; DST[i] = SRC1[i]+SRC2[i] for all i
0x01 1 DP3 Computes dot product on 3-component vectors; DST = SRC1.SRC2
0x02 1 DP4 Computes dot product on 4-component vectors; DST = SRC1.SRC2
0x03 1 DPH Computes dot product on a 3-component vector with 1.0 appended to it and a 4-component vector; DST = SRC1.SRC2 (with SRC1 homogenous)
0x04 1 DST Equivalent to Microsoft's dst instruction: DST = {1, SRC1[1]*SRC2[1], SRC1[2], SRC2[3]}
0x05 1u EX2 Computes SRC1's first component exponent with base 2; DST[i] = EXP2(SRC1[0]) for all i
0x06 1u LG2 Computes SRC1's first component logarithm with base 2; DST[i] = LOG2(SRC1[0]) for all i
0x07 1u LITP Partial lighting computation, may be used in conjunction with EX2, LG2, etc to compute the vertex lighting coefficients. See the Microsoft and ARB docs for more information on how to implement the full lit function; DST = {max(src.x, 0), max(min(src.y, 127.9961), -127.9961), 0, max(src.w, 0)} and it sets the cmp.x and cmp.y flags based on if the respective src.x and src.w components are >= 0.
0x08 1 MUL Multiplies two vectors component by component; DST[i] = SRC1[i].SRC2[i] for all i
0x09 1 SGE Sets output if SRC1 is greater than or equal to SRC2; DST[i] = (SRC1[i] >= SRC2[i]) ? 1.0 : 0.0 for all i
0x0A 1 SLT Sets output if SRC1 is strictly less than SRC2; DST[i] = (SRC1[i] < SRC2[i]) ? 1.0 : 0.0 for all i
0x0B 1u FLR Computes SRC1's floor component by component; DST[i] = FLOOR(SRC1[i]) for all i
0x0C 1 MAX Takes the max of two vectors, component by component; DST[i] = MAX(SRC1[i], SRC2[i]) for all i
0x0D 1 MIN Takes the min of two vectors, component by component; DST[i] = MIN(SRC1[i], SRC2[i]) for all i
0x0E 1u RCP Computes the reciprocal of the vector's first component; DST[i] = 1/SRC1[0] for all i
0x0F 1u RSQ Computes the reciprocal of the square root of the vector's first component; DST[i] = 1/sqrt(SRC1[0]) for all i
0x10 ? ??? ?
0x11 ? ??? ?
0x12 1u MOVA Move to address register; Casts the float value given by SRC1 to an integer (truncating the fractional part) and assigns the result to (a0.x, a0.y, _, _), respecting the destination component mask.
0x13 1u MOV Moves value from one register to another; DST = SRC1.
0x14 ? ??? ?
0x15 ? ??? ?
0x16 ? ??? ?
0x17 ? ??? ?
0x18 1i DPHI Computes dot product on a 3-component vector with 1.0 appended to it and a 4-component vector; DST = SRC1.SRC2 (with SRC1 homogenous)
0x19 1i DSTI DST with sources swapped.
0x1A 1i SGEI Sets output if SRC1 is greater than or equal to SRC2; DST[i] = (SRC1[i] >= SRC2[i]) ? 1.0 : 0.0 for all i
0x1B 1i SLTI Sets output if SRC1 is strictly less than SRC2; DST[i] = (SRC1[i] < SRC2[i]) ? 1.0 : 0.0 for all i
0x1C ? ??? ?
0x1D ? ??? ?
0x1E ? ??? ?
0x1F ? ??? ?
0x20 0 BREAK Breaks out of LOOP block; do not use while in nested IF/CALL block inside LOOP block.
0x21 0 NOP Does literally nothing.
0x22 0 END Signals the shader unit that processing for this vertex/primitive is done.
0x23 2 BREAKC If condition (see below for details) is true, then breaks out of LOOP block.
0x24 2 CALL Jumps to DST and executes instructions until it reaches DST+NUM instructions
0x25 2 CALLC If condition (see below for details) is true, then jumps to DST and executes instructions until it reaches DST+NUM instructions, else does nothing.
0x26 3 CALLU Jumps to DST and executes instructions until it reaches DST+NUM instructions if BOOL is true
0x27 3 IFU If condition BOOL is true, then executes instructions until DST, then jumps to DST+NUM; else, jumps to DST.
0x28 2 IFC If condition (see below for details) is true, then executes instructions until DST, then jumps to DST+NUM; else, jumps to DST
0x29 3 LOOP Loops over the code between itself and DST (inclusive), performing INT.x+1 iterations in total. First, aL is initialized to INT.y. After each iteration, aL is incremented by INT.z.
0x2A 0 (no param) EMIT (geometry shader only) Emits a vertex (and primitive if FLAG_PRIMEMIT was set in the corresponding SETEMIT). SETEMIT must be called before this.
0x2B 4 SETEMIT (geometry shader only) Sets VTXID, FLAG_WINDING and FLAG_PRIMEMIT for the next EMIT instruction. VTXID is the ID of the vertex about to be emitted within the primitive, while FLAG_PRIMEMIT is zero if we are just emitting a single vertex and non-zero if are emitting a vertex and primitive simultaneously. FLAG_WINDING controls the output primitive's winding. Note that the output vertex buffer (which holds 4 vertices) is not cleared when the primitive is emitted, meaning that vertices from the previous primitive can be reused for the current one. (this is still a working hypothesis and unconfirmed)
0x2C 2 JMPC If condition (see below for details) is true, then jumps to DST, else does nothing.
0x2D 3 JMPU If condition BOOL is true, then jumps to DST, else does nothing. Having bit 0 of NUM = 1 will invert the test, jumping if BOOL is false instead.
0x2E-0x2F 1c CMP Sets booleans cmp.x and cmp.y based on the operand's x and y components and the CMPX and CMPY comparison operators respectively. See below for details about operators. It's unknown whether CMP respects the destination component mask or not.
0x30-0x37 5i MADI Multiplies two vectors and adds a third one component by component; DST[i] = SRC3[i] + SRC2[i].SRC1[i] for all i; this is not an FMA, the intermediate result is rounded
0x38-0x3F 5 MAD Multiplies two vectors and adds a third one component by component; DST[i] = SRC3[i] + SRC2[i].SRC1[i] for all i; this is not an FMA, the intermediate result is rounded

Operand descriptors[edit]

Sizes below are in bits, not bytes.

Offset Size Description
0x0 0x4 Destination component mask. Bit 3 = x, 2 = y, 1 = z, 0 = w.
0x4 0x1 Source 1 negation bit
0x5 0x8 Source 1 component selector
0xD 0x1 Source 2 negation bit
0xE 0x8 Source 2 component selector
0x16 0x1 Source 3 negation bit
0x17 0x8 Source 3 component selector

Component selector :

Offset Size Description
0x0 0x2 Component 3 value
0x2 0x2 Component 2 value
0x4 0x2 Component 1 value
0x6 0x2 Component 0 value
Value Component
0x0 x
0x1 y
0x2 z
0x3 w

The component selector enables swizzling. For example, component selector 0x1B is equivalent to .xyzw, while 0x55 is equivalent to .yyyy.

Depending on the current shader opcode, source components are disabled implicitly by setting the destination component mask. For example, ADD o0.xy, r0.xyzw, r1.xyzw will not make use of r0's or r1's z/w components, while DP4 o0.xy, r0.xyzw, r1.xyzw will use all input components regardless of the used destination component mask.

Relative addressing[edit]

IDX raw value Register name
0x0 None
0x1 a0.x
0x2 a0.y
0x3 aL

There are 3 address registers: a0.x, a0.y and aL (loop counter). For format 1 instructions, when IDX != 0, the value of the corresponding address register is added to SRC1's value. For example, if IDX = 2, a0.y = 3 and SRC1 = c8, then instead SRC1+a0.y = c11 will be used for the instruction. It is only possible to use address registers on constant registers, attempting to use them on input attribute or temporary registers results in the address register being ignored (i.e. read as zero).

a0.x and a0.y are set manually through the MOVA instruction by rounding a float value to integer precision. Hence, they may take negative values. The way out-of-bounds values behave when reading uniforms is as follows:

  • If the offset is out of byte bounds (less than -128 or greater than 127), the offset is not applied (treated as 0).
  • The offset is added to the constant register index and masked by 0x7F.
  • If the resulting index is greater than 95, the result is (1, 1, 1, 1).
  • Otherwise, the result is the value at the indexed constant register.

aL can only be set indirectly by the LOOP instruction. It is still accessible and valid after exiting a LOOP block, though.

Comparison operator[edit]

CMPX/CMPY raw value Operator name Expression
0x0 EQ src1 == src2
0x1 NE src1 != src2
0x2 LT src1 < src2
0x3 LE src1 <= src2
0x4 GT src1 > src2
0x5 GE src1 >= src2
0x6 ?? true ?
0x7 ?? true ?

6 and 7 seem to always return true.

Conditions[edit]

A number of format 2 instructions are executed conditionally. These conditions are based on two boolean registers which can be set with CMP : cmp.x and cmp.y.

Conditional instructions include 3 parameters : CONDOP, REFX and REFY. REFX and REFY are reference values which are tested for equality against cmp.x and cmp.y, respectively. CONDOP describes how the final truth value is constructed from the results of the two tests. There are four conditional expression formats :

CONDOP raw value Expression Description
0x0 cmp.x == REFX || cmp.y == REFY OR
0x1 cmp.x == REFX && cmp.y == REFY AND
0x2 cmp.x == REFX X
0x3 cmp.y == REFY Y

Registers[edit]

Name Format Type Access Written by Description
v0-v15 vector float Read only Application/Vertex-stream Input registers.
o0-o15 vector float Write only Vertex shader Output registers.
r0-r15 vector float Read/Write Vertex shader Temporary registers.
c0-c95 vector float Read only Application/Vertex-stream Floating-point Constant registers.
i0-i3 vector integer Read only Application Integer Constant registers. (special purpose)
b0-b15 scalar boolean Read only Application Boolean Constant registers. (special purpose)
a0.x & a0.y scalar integer Use/Write Vertex shader Address registers.
aL scalar integer Use Vertex shader Loop count register.

Input attribute registers store the per-vertex data given by the CPU and hence are read-only.

Output registers hold the data to be passed to the later GPU stages and are write-only. Each of the output register is assigned a semantic by setting the corresponding GPU_Internal_Registers. Output registers o7-o15 are only available in vertex shaders. Keep in mind that writing to the same output register/component more than once appears appears to cause problems (e.g. GPU hangs).

Temporary registers can be used for intermediate calculations and can be both read and written.

Constant registers hold data uploaded by the application which remain constant throughout all processed vertices. There are 96 float[4] constant registers (c0-c95), eight boolean constant registers (b0-b7), and four int[4] constant registers (i0-i3). Many shader instructions which take float arguments can only provide the full 7 bits for one SRC operand. All other source operands can only be used to refer to input attributes or temporary registers and cannot be passed Floating-point Constant registers.

Address registers and the Loop count register can be used to to provide relative addressing for the designated SRC operand. For more information, see the section on relative addressing.

DST mapping :

DST raw value Register name Description
0x0-0xF o0-o15 Output registers.
0x10-0x1F r0-r15 Temporary registers.

SRC mapping :

SRC raw value Register name Description
0x0-0xF v0-v15 Input attribute registers.
0x10-0x1F r0-r15 Temporary registers.
0x20-0x7F c0-c95 Constant registers.

Floating-Point Behavior[edit]

The PICA200 is not IEEE-compliant. It has positive and negative infinities and NaN, but does not seem to have negative 0. Input and output subnormals are flushed to +0. The internal floating point format seems to be the same as used in shader binaries: 1 sign bit, 7 exponent bits, 16 (explicit) mantissa bits. Several instructions also have behavior that differs from the IEEE functions. Here are the results from some tests done on hardware (s = largest subnormal, n = smallest positive normal):

Computation Result Notes
inf * 0 0 Including inside MUL, MAD, DP4, etc.
NaN * 0 NaN
+inf - +inf NaN Indicates +inf is real inf, not FLT_MAX
rsq(rcp(-inf)) +inf Indicates that there isn't -0.0.
rcp(-0) +inf no -0 so differs from IEEE where rcp(-0) = -inf
rcp(0) +inf
rcp(+inf) 0
rcp(NaN) NaN
rsq(-0) +inf no -0 so differs from IEEE where rsq(-0) = -inf
rsq(-2) NaN
rsq(+inf) 0
rsq(-inf) NaN
rsq(NaN) NaN
max(0, +inf) +inf
max(0, -inf) -inf
max(0, NaN) NaN max violates IEEE but match GLSL spec
max(NaN, 0) 0
max(-inf, +inf) +inf
min(0, +inf) 0
min(0, -inf) -inf
min(0, NaN) NaN min violates IEEE but match GLSL spec
min(NaN, 0) 0
min(-inf, +inf) -inf
cmp(s, 0) false cmp does not flush input subnormals
max(s, 0) s max does not flush input or output subnormals
mul(s, 2) 0 input subnormals are flushed in arithmetic instructions
mul(n, 0.5) 0 output subnormals are flushed in arithmetic instructions

1.0 can be multiplied 63 times by 0.5 until the result compares equal zero. This is consistent with a 7-bit exponent and output subnormal flushing.

Control Flow[edit]

Control flow is implemented using four independent stacks:

  • 4-deep CALL stack
  • 8-deep IF stack
  • 4-deep LOOP stack

All stacks are initially empty. After every instruction but before JMP takes effect, the PC is incremented and a copy is sent to each stack. Each stack is checked against its copy of the PC. If an entry is popped from the stack, the copied PC is updated and used for the next check of this stack, although the IF/LOOP stacks can each only pop one entry per instruction, whereas the CALL stack is checked again until it doesn't match or the stack is empty. The updated PC copy with the highest priority wins: LOOP (highest), IF, CALL, JMP, original PC (lowest).

Special cases:

  • JMP overwrites the PC *after* the stacks checks (and only if no stack was popped).
  • Executing a BREAK on an empty LOOP stack hangs the GPU.
  • A stack overflow discards the oldest element, so you could think of it as a queue or a ring buffer.
  • If the CALL stack is popped four times in a row, the fourth update to its copy of the PC is missed (the third PC update will be propagated). Probably a hardware bug.