X-Git-Url: https://git.libre-soc.org/?a=blobdiff_plain;f=openpower%2Fsv%2Fvector_ops.mdwn;h=1a69d92070cc2980debbb5d6aa30e5ee0487c546;hb=88a823723277b6f79040042f9f2c1a00b33d7f8d;hp=6a80ee3b87d8519693790cf69a78577c550d70c1;hpb=2e37f5cdb9e1f0e128eeb7c50a976c51717b1f06;p=libreriscv.git

diff --git a/openpower/sv/vector_ops.mdwn b/openpower/sv/vector_ops.mdwn
index 6a80ee3b8..1a69d9207 100644
--- a/openpower/sv/vector_ops.mdwn
+++ b/openpower/sv/vector_ops.mdwn
@@ -1,11 +1,17 @@
+[[!tag standards]]
+
 # SV Vector Operations.
 
-The core OpenPOWER ISA was designed as scalar: SV provides a level of abstraction to add variable-length element-independent parallelism. However, certain classes of instructions only make sense in a Vector context: AVC512 conflictd for example.  This section includes such examples.  Many of them are from the RISC-V Vector ISA (with thanks to the efforts of RVV's contributors)
+TODO merge old standards page [[simple_v_extension/vector_ops/]]
+
+The core OpenPOWER ISA was designed as scalar: SV provides a level of abstraction to add variable-length element-independent parallelism. However, certain classes of instructions only make sense in a Vector context: AVX512 conflictd for example.  This section includes such examples.  Many of them are from the RISC-V Vector ISA (with thanks to the efforts of RVV's contributors)
+
+Notes:
 
-However some of these actually could be added to a scalar ISA as bitmanipulation instructions.  These are separated out into their own section.
-Instructions suited to 3D GPU workloads (dotproduct, crossproduct, normalise) are out of scope: this document is for more general-purpose instructions that underpin and are critical to general-purpose Vector workloads (including GPU and VPU)
+* Some of these actually could be added to a scalar ISA as bitmanipulation instructions.  These are separated out into their own section.
+* Instructions suited to 3D GPU workloads (dotproduct, crossproduct, normalise) are out of scope: this document is for more general-purpose instructions that underpin and are critical to general-purpose Vector workloads (including GPU and VPU)
+* Instructions related to the adaptation of CRs for use as predicate masks are covered separately, by crweird operations.  See [[sv/cr_int_predication]].
 
-.
 Links:
 
 * <https://github.com/riscv/riscv-v-spec/blob/master/v-spec.adoc#vector-register-gather-instructions>
@@ -13,12 +19,14 @@ Links:
 * <https://bugs.libre-soc.org/show_bug.cgi?id=213>
 * <https://bugs.libre-soc.org/show_bug.cgi?id=142> specialist vector ops
  out of scope for this document
+* [[simple_v_extension/specification/bitmanip]] previous version,
+  contains pseudocode for sof, sif, sbf
 
 # Vector
 
 ## conflictd
 
-This is based on the AVX512 conflict detection instruction.  Internally the logic is used to detect address conflicts in LD/ST operations.  Two arrays of indices are given.
+This is based on the AVX512 conflict detection instruction.  Internally the logic is used to detect address conflicts in multi-issue LD/ST operations.  Two arrays of values are given: the indices are compared and duplicates reported in a triangular fashion.  the instruction may be used for histograms (computed in parallel)
 
     input = [100, 100,   3, 100,   5, 100, 100,   3]
     conflict result = [
@@ -32,15 +40,22 @@ This is based on the AVX512 conflict detection instruction.  Internally the logi
          0b00000100     // 3 is present on #2
     ]
 
+Pseudocode:
+
+    for i in range(VL):
+        for j in range(1, i):
+            if src1[i] == src2[j]:
+                result[j] |= 1<<i
+
 ## iota
 
-Based on RVV vmiota.  vmiota may be viewed as a cumulative variant of cntlz, where instead of stopping at the first zero with a count to produce a single scalar result, the process continues on, producing another element at the next encounter of a 1.
+Based on RVV vmiota.  vmiota may be viewed as a cumulative variant of popcount, generating multiple results.  successive iterations include more and more bits of the bitstream being tested.
 
-The viota.m instruction reads a source vector mask register and writes to each element of the destination vector register group the sum of all the bits of elements in the mask register whose index is less than the element, e.g., a parallel prefix sum of the mask values.
+When masked, only the bits not masked out are included in the count process.
 
-This instruction can be masked, in which case only the enabled elements contribute to the sum and only the enabled elements are written.
+    viota RT/v, RA, RB
 
-    viota.m vd, vs2, vm
+Note that when RA=0 this indicates to test against all 1s, resulting in the instruction generating a vector sequence [0, 1, 2... VL-1]. This will be equivalent to RVV vid.m which is a pseudo-op, here (RA=0).
 
 Example
 
@@ -56,11 +71,31 @@ Example
                        viota.m v4, v2, v0.t # Masked
      1 1 1 5 1 7 1 0   v4 results
 
-The result value is zero-extended to fill the destination element if SEW is wider than the result. If the result value would overflow the destination SEW, the least-significant SEW bits are retained.
+     def iota(RT, RA, RB): 
+        mask = RB ? iregs[RB] : 0b111111...1
+        val = RA ? iregs[RA] : 0b111111...1
+        for i in range(VL):
+            if RA.scalar:
+            testmask = (1<<i)-1 # only count below
+            to_test = val & testmask & mask
+            iregs[RT+i] = popcount(to_test)
 
-Traps on viota.m are always reported with a vstart of 0, and execution is always restarted from the beginning when resuming after a trap handler. An illegal instruction exception is raised if vstart is non-zero.
+a Vector CR-based version of the same, due to CRs being used for predication. This would use the same testing mechanism as branch: BO[0:2]
+where bit 2 is inv, bits 0:1 select the bit of the CR.
 
+     def test_CR_bit(CR, BO):
+         return CR[BO[0:1]] == BO[2]
 
+     def iotacr(RT, BA, BO): 
+        mask = get_src_predicate()
+        count = 0
+        for i in range(VL):
+            if mask & (1<<i) == 0: continue
+            iregs[RT+i] = count
+            if test_CR_bit(CR[i+BA], BO):
+                 count += 1
+
+the variant of iotacr which is vidcr, this is not appropriate to have BA=0, plus, it is pointless to have it anyway.  The integer version covers it, by not reading the int regfile at all.
 
 # Scalar
 
@@ -93,6 +128,27 @@ Example
 
 The vmsbf.m instruction takes a mask register as input and writes results to a mask register. The instruction writes a 1 to all active mask elements before the first source element that is a 1, then writes a 0 to that element and all following active elements. If there is no set bit in the source vector, then all active elements in the destination are written with a 1.
 
+pseudocode:
+
+    def sbf(rd, rs1, rs2):
+        rd = 0
+        # start setting if no predicate or if 1st predicate bit set
+        setting_mode = rs2 == x0 or (regs[rs2] & 1)
+        while i < XLEN:
+            bit = 1<<i
+            if rs2 != x0 and (regs[rs2] & bit):
+                # reset searching
+                setting_mode = False
+            if setting_mode:
+                if regs[rs1] & bit: # found a bit in rs1: stop setting rd
+                    setting_mode = False
+                else:
+                    regs[rd] |= bit
+            else if rs2 != x0: # searching mode
+                if (regs[rs2] & bit):
+                    setting_mode = True # back into "setting" mode
+            i += 1
+
 ## sifm
 
 The vector mask set-including-first instruction is similar to set-before-first, except it also includes the element with a set bit.
@@ -116,13 +172,44 @@ The vector mask set-including-first instruction is similar to set-before-first,
                        vmsif.m v2, v3, v0.t
      1 1 x x x x 1 1   v2 contents
 
+Pseudo-code:
+
+    def sif(rd, rs1, rs2):
+        rd = 0
+        setting_mode = rs2 == x0 or (regs[rs2] & 1)
+
+        while i < XLEN:
+            bit = 1<<i
+
+            # only reenable when predicate in use, and bit valid
+            if !setting_mode && rs2 != x0:
+                if (regs[rs2] & bit):
+                    # back into "setting" mode
+                    setting_mode = True
+
+            # skipping mode
+            if !setting_mode:
+                # skip any more 1s
+                if regs[rs1] & bit == 1:
+                    i += 1
+                    continue
+
+            # setting mode, search for 1
+            regs[rd] |= bit # always set during search
+            if regs[rs1] & bit: # found a bit in rs1:
+                setting_mode = False
+                # next loop starts skipping
+
+            i += 1
+
+
 ## vmsof
 
 The vector mask set-only-first instruction is similar to set-before-first, except it only sets the first element with a bit set, if any.
 
     sofm RT, RA, RB
 
- # Example
+Example
 
      7 6 5 4 3 2 1 0   Bit number
 
@@ -139,3 +226,70 @@ The vector mask set-only-first instruction is similar to set-before-first, excep
                        vmsof.m v2, v3, v0.t
      0 1 x x x x 0 0   v2 content
 
+Pseudo-code:
+
+    def sof(rd, rs1, rs2):
+        rd = 0
+        setting_mode = rs2 == x0 or (regs[rs2] & 1)
+
+        while i < XLEN:
+            bit = 1<<i
+
+            # only reenable when predicate in use, and bit valid
+            if !setting_mode && rs2 != x0:
+                if (regs[rs2] & bit):
+                    # back into "setting" mode
+                    setting_mode = True
+
+            # skipping mode
+            if !setting_mode:
+                # skip any more 1s
+                if regs[rs1] & bit == 1:
+                    i += 1
+                    continue
+
+            # setting mode, search for 1
+            if regs[rs1] & bit: # found a bit in rs1:
+                regs[rd] |= bit # only set when search succeeds
+                setting_mode = False
+                # next loop starts skipping
+
+            i += 1
+
+# Carry-lookahead
+
+used not just for carry lookahead, also a special type of predication mask operation.
+
+* <https://www.geeksforgeeks.org/carry-look-ahead-adder/>
+* <https://media.geeksforgeeks.org/wp-content/uploads/digital_Logic6.png>
+* <https://electronics.stackexchange.com/questions/20085/whats-the-difference-with-carry-look-ahead-generator-block-carry-look-ahead-ge>
+* <https://i.stack.imgur.com/QSLKY.png>
+* <https://stackoverflow.com/questions/27971757/big-integer-addition-code>
+  `((P|G)+G)^P`
+* <https://en.m.wikipedia.org/wiki/Carry-lookahead_adder>
+
+```
+     P = (A | B) & Ci
+     G = (A & B)
+```
+
+Stackoverflow algorithm `((P|G)+G)^P` works on the cumulated bits of P and G from associated vector units (P and G are integers here).  The result of the algorithm is the new carry-in which already includes ripple, one bit of carry per element.
+
+```
+    At each id, compute C[id] = A[id]+B[id]+0
+    Get G[id] = C[id] > radix -1
+    Get P[id] = C[id] == radix-1
+    Join all P[id] together, likewise G[id]
+    Compute newC = ((P|G)+G)^P
+    result[id] = (C[id] + newC[id]) % radix
+```   
+
+two versions: scalar int version and CR based version.
+
+scalar int version acts as a scalar carry-propagate, reading XER.CA as input, P and G as regs, and taking a radix argument.  the end bits go into XER.CA and CR0.ge
+
+vector version takes CR0.so as carry in, stores in CR0.so and CR.ge end bits.
+
+if zero (no propagation) then CR0.eq is zero
+
+CR based version, TODO.