Difference between revisions of "Instruction Set/brtr"

From Mill Computing Wiki
Jump to: navigation, search
(Created page with "{{DISPLAYTITLE:brtr}} <div style="font-size:80%;line-height:90%;margin-bottom:2em">realizing  flow stream Decode|flow block...")
 
Line 4:Line 4:
 
</div>
 
</div>
  
−
branch
+
Branch on true predicate.
 +
There can be several conditionless branches in an [[EBB]] and even in the same operation, which are all processes in parallel, but the first successful in the lowest slot wins.
 +
 
 +
The targets in branches, whether literal or from a belt operand, are always relative to the  [[Registers|EBB entry point]]. The optional delay serves to synchronize with operations that need to finish before control is transferred to the next EBB. This is particularly important for the predicates to examine for the branch, the value of which is examined after the delay.
 +
 
 +
The branch not taken case is more efficient and faster, i.e. the compiler takes care to schedule the conditional branches with their more likely case not to be taken, to achieve the longest possible code sequences without control transfers.
 +
 
 +
<b>related operations:</b> [[Instruction_Set/br|br]], [[Instruction_Set/brfl|brfl]]
 +
 
 
----
 
----
 
<code style="font-size:130%"><b style="color:#050">brtr</b>(<span style="color:#009">[[Domains#pred|pred]]</span> <span title="late-evaluated 1-bit predicate from belt">q</span>, <span style="color:#009">[[Domains#p|p]]</span> <span title="belt operand from opsWindow">target</span>)</code>
 
<code style="font-size:130%"><b style="color:#050">brtr</b>(<span style="color:#009">[[Domains#pred|pred]]</span> <span title="late-evaluated 1-bit predicate from belt">q</span>, <span style="color:#009">[[Domains#p|p]]</span> <span title="belt operand from opsWindow">target</span>)</code>
Line 14:Line 22:
 
! [[Cores|Core]] || [[Slot|In Slots]]|| [[Latency|Latencies]]
 
! [[Cores|Core]] || [[Slot|In Slots]]|| [[Latency|Latencies]]
 
|-
 
|-
−
| [[Cores/Tin/Encoding#227|Tin]] || F0 || 1
+
| [[Cores/Tin/Encoding#brtr|Tin]] || F0 || 1
 
|-
 
|-
−
| [[Cores/Copper/Encoding#227|Copper]] || F0 F1 || 1
+
| [[Cores/Copper/Encoding#brtr|Copper]] || F0 F1 || 1
 
|-
 
|-
−
| [[Cores/Silver/Encoding#227|Silver]] || F0 F1 F2 || 1
+
| [[Cores/Silver/Encoding#brtr|Silver]] || F0 F1 F2 || 1
 
|-
 
|-
−
| [[Cores/Gold/Encoding#227|Gold]] || F0 F1 F2 F3 || 1
+
| [[Cores/Gold/Encoding#brtr|Gold]] || F0 F1 F2 F3 || 1
 
|-
 
|-
−
| [[Cores/Decimal8/Encoding#227|Decimal8]] || F0 F1 F2 || 1
+
| [[Cores/Decimal8/Encoding#brtr|Decimal8]] || F0 F1 F2 || 1
 
|-
 
|-
−
| [[Cores/Decimal16/Encoding#227|Decimal16]] || F0 F1 F2 || 1
+
| [[Cores/Decimal16/Encoding#brtr|Decimal16]] || F0 F1 F2 || 1
 
|}
 
|}
  
Line 37:Line 45:
 
! [[Cores|Core]] || [[Slot|In Slots]]|| [[Latency|Latencies]]
 
! [[Cores|Core]] || [[Slot|In Slots]]|| [[Latency|Latencies]]
 
|-
 
|-
−
| [[Cores/Tin/Encoding#228|Tin]] || F0 || 1
+
| [[Cores/Tin/Encoding#brtr|Tin]] || F0 || 1
 
|-
 
|-
−
| [[Cores/Copper/Encoding#228|Copper]] || F0 F1 || 1
+
| [[Cores/Copper/Encoding#brtr|Copper]] || F0 F1 || 1
 
|-
 
|-
−
| [[Cores/Silver/Encoding#228|Silver]] || F0 F1 F2 || 1
+
| [[Cores/Silver/Encoding#brtr|Silver]] || F0 F1 F2 || 1
 
|-
 
|-
−
| [[Cores/Gold/Encoding#228|Gold]] || F0 F1 F2 F3 || 1
+
| [[Cores/Gold/Encoding#brtr|Gold]] || F0 F1 F2 F3 || 1
 
|-
 
|-
−
| [[Cores/Decimal8/Encoding#228|Decimal8]] || F0 F1 F2 || 1
+
| [[Cores/Decimal8/Encoding#brtr|Decimal8]] || F0 F1 F2 || 1
 
|-
 
|-
−
| [[Cores/Decimal16/Encoding#228|Decimal16]] || F0 F1 F2 || 1
+
| [[Cores/Decimal16/Encoding#brtr|Decimal16]] || F0 F1 F2 || 1
 
|}
 
|}
  
Line 59:Line 67:
 
! [[Cores|Core]] || [[Slot|In Slots]]|| [[Latency|Latencies]]
 
! [[Cores|Core]] || [[Slot|In Slots]]|| [[Latency|Latencies]]
 
|-
 
|-
−
| [[Cores/Tin/Encoding#226|Tin]] || F0 || 1
+
| [[Cores/Tin/Encoding#brtr|Tin]] || F0 || 1
 
|-
 
|-
−
| [[Cores/Copper/Encoding#226|Copper]] || F0 F1 || 1
+
| [[Cores/Copper/Encoding#brtr|Copper]] || F0 F1 || 1
 
|-
 
|-
−
| [[Cores/Silver/Encoding#226|Silver]] || F0 F1 F2 || 1
+
| [[Cores/Silver/Encoding#brtr|Silver]] || F0 F1 F2 || 1
 
|-
 
|-
−
| [[Cores/Gold/Encoding#226|Gold]] || F0 F1 F2 F3 || 1
+
| [[Cores/Gold/Encoding#brtr|Gold]] || F0 F1 F2 F3 || 1
 
|-
 
|-
−
| [[Cores/Decimal8/Encoding#226|Decimal8]] || F0 F1 F2 || 1
+
| [[Cores/Decimal8/Encoding#brtr|Decimal8]] || F0 F1 F2 || 1
 
|-
 
|-
−
| [[Cores/Decimal16/Encoding#226|Decimal16]] || F0 F1 F2 || 1
+
| [[Cores/Decimal16/Encoding#brtr|Decimal16]] || F0 F1 F2 || 1
 
|}
 
|}
  
Line 82:Line 90:
 
! [[Cores|Core]] || [[Slot|In Slots]]|| [[Latency|Latencies]]
 
! [[Cores|Core]] || [[Slot|In Slots]]|| [[Latency|Latencies]]
 
|-
 
|-
−
| [[Cores/Tin/Encoding#225|Tin]] || F0 || 1
+
| [[Cores/Tin/Encoding#brtr|Tin]] || F0 || 1
 
|-
 
|-
−
| [[Cores/Copper/Encoding#225|Copper]] || F0 F1 || 1
+
| [[Cores/Copper/Encoding#brtr|Copper]] || F0 F1 || 1
 
|-
 
|-
−
| [[Cores/Silver/Encoding#225|Silver]] || F0 F1 F2 || 1
+
| [[Cores/Silver/Encoding#brtr|Silver]] || F0 F1 F2 || 1
 
|-
 
|-
−
| [[Cores/Gold/Encoding#225|Gold]] || F0 F1 F2 F3 || 1
+
| [[Cores/Gold/Encoding#brtr|Gold]] || F0 F1 F2 F3 || 1
 
|-
 
|-
−
| [[Cores/Decimal8/Encoding#225|Decimal8]] || F0 F1 F2 || 1
+
| [[Cores/Decimal8/Encoding#brtr|Decimal8]] || F0 F1 F2 || 1
 
|-
 
|-
−
| [[Cores/Decimal16/Encoding#225|Decimal16]] || F0 F1 F2 || 1
+
| [[Cores/Decimal16/Encoding#brtr|Decimal16]] || F0 F1 F2 || 1
 
|}
 
|}
 +
 +
 +
[[Instruction_Set|Instruction Set, alphabetical]], [[Instruction Set by Category]], [http://millcomputing.com/instructions.html?collapse=7#ops Instruction Set, sortable, filterable]

Revision as of 02:36, 16 December 2014

realizing  flow stream  flow block  transfer phase   operation  

native on: all

Branch on true predicate. There can be several conditionless branches in an EBB and even in the same operation, which are all processes in parallel, but the first successful in the lowest slot wins.

The targets in branches, whether literal or from a belt operand, are always relative to the EBB entry point. The optional delay serves to synchronize with operations that need to finish before control is transferred to the next EBB. This is particularly important for the predicates to examine for the branch, the value of which is examined after the delay.

The branch not taken case is more efficient and faster, i.e. the compiler takes care to schedule the conditional branches with their more likely case not to be taken, to achieve the longest possible code sequences without control transfers.

related operations: br, brfl

brtr(pred q, p target)

operands: like Inv :


Core In Slots Latencies
Tin F0 1
Copper F0 F1 1
Silver F0 F1 F2 1
Gold F0 F1 F2 F3 1
Decimal8 F0 F1 F2 1
Decimal16 F0 F1 F2 1

brtr(pred q, p target, lit delay)

operands: like Inv :


Core In Slots Latencies
Tin F0 1
Copper F0 F1 1
Silver F0 F1 F2 1
Gold F0 F1 F2 F3 1
Decimal8 F0 F1 F2 1
Decimal16 F0 F1 F2 1

brtr(pred q, lbl target)

operands: like Inv :


Core In Slots Latencies
Tin F0 1
Copper F0 F1 1
Silver F0 F1 F2 1
Gold F0 F1 F2 F3 1
Decimal8 F0 F1 F2 1
Decimal16 F0 F1 F2 1

brtr(pred q, lbl target, lit delay)

operands: like Inv :


Core In Slots Latencies
Tin F0 1
Copper F0 F1 1
Silver F0 F1 F2 1
Gold F0 F1 F2 F3 1
Decimal8 F0 F1 F2 1
Decimal16 F0 F1 F2 1


Instruction Set, alphabetical, Instruction Set by Category, Instruction Set, sortable, filterable

Retrieved from "http://millcomputing.com/w/index.php?title=Instruction_Set/brtr&oldid=1572"