Minor improvements to docs (#46)

Author: giordano

docs/src/compiler.md

@@ -37,7 +37,7 @@ Inside codelets defined with [`@codelet`](@ref) all calls to random functions
 * `randn(Float32)`
 
 result to call to corresponding IPU builtins for [random number generation](https://docs.graphcore.ai/projects/poplar-api/en/latest/ipu_intrinsics/ipu_builtins.html#random-number-generation).
-The uniformly distributed numbers follow the general semantic of the Julia function `rand` (numbers uniformely distributed in the $[0, 1)$ range), while the normally distributed numbers have the properties described in the Poplar SDK documentation (numbers are in the range $[-5-13/16, 5+13/16]$).
+The uniformly distributed numbers follow the general semantic of the Julia function `rand` (floating point numbers are uniformely distributed in the $[0, 1)$ range), while the normally distributed numbers have the properties described in the Poplar SDK documentation (numbers are in the range $[-93/16, 93/16]$).
 
 Additionally, you can use the [IPU builtins](https://docs.graphcore.ai/projects/poplar-api/en/latest/ipu_intrinsics/ipu_builtins.html) listed below.
 
@@ -87,7 +87,7 @@ The printing macros `@ipucycles` and `@ipushowcycles` can be made completely no-
 
 If your kernel references a non-constant (`const`) global variable, the generated code will result in a reference to a memory address on the host, and this will fatally fail at runtime because programs running on the IPU don't have access to the host memory.
 Constant variables are not affected by this problem because their values are inlined when the function is compiled.
-If you can't or don't want to make a variable constant you can interpolate its value with a top-level `@eval` when defining the codelet.
+If you can't or don't want to make a variable constant you can interpolate its value with a top-level [`@eval`](https://docs.julialang.org/en/v1/base/base/#Base.@eval) when defining the codelet.
 For example:
 
 ```julia

docs/src/index.md

@@ -30,7 +30,7 @@ Other versions of the Poplar SDK are not currently supported.
     * the Poplar SDK version 2.2 uses LLVM 13, which is available in Julia v1.8;
     * the Poplar SDK versions 2.3-2.5 use LLVM 14, which is available in Julia v1.9;
     * the Poplar SDK versions 2.6-3.2 use LLVM 15, which is available in Julia v1.10;
-    * the Poplar SDK version 3.3 use LLVM 16, which as of 2023-07-05 is not available in any Julia version.
+    * the Poplar SDK version 3.3 uses LLVM 16, which is available in Julia v1.11.
 
 ## Installation
 
@@ -72,5 +72,6 @@ These approaches are exploratory of the functionalities, and are often limited i
 Here is some material that you may find useful for learning more about Julia on the IPU and trying it out yourself:
 
 * [Pluto notebook](https://giordano.github.io/blog/2023-07-20-julia-ipu/) of presentation given at Graphcore and at JuliaCon in July 2023
+* Talk "[Julia meets the Intelligence Processing Unit](https://www.youtube.com/watch?v=-fxB0kmcCVE)" at JuliaCon 2023
 * [JuliaIpuDemo](https://github.com/JuliaIPU/JuliaIpuDemo), repository with instructions for running some Jupyter notebooks on Paperspace cloud.
   This service offers also IPU time for free, these sessions are limited to 6 hours each and one at the time, but you can run as many as you want.

src/compiler/output.jl

@@ -53,7 +53,7 @@ to match, eg. printing a 64-bit Julia integer requires the `%ld` formatting stri
 More user-friendly versions of this macro are [`@ipuprint`](@ref), [`@ipuprintln`](@ref).
 See also [`@ipushow`](@ref), which is built on top of `@ipuprintf` functionalities.
 
-Printing can be completely disabled by setting
+Printing can be completely disabled by setting [`IPUCompiler.DISABLE_PRINT`](@ref):
 ```julia
 $(@__MODULE__).DISABLE_PRINT[] = true
 ```
@@ -241,7 +241,7 @@ Limited string interpolation is also possible:
     @ipuprint "Hello, World \$(42)\\n"
 ```
 
-Printing can be completely disabled by setting
+Printing can be completely disabled by setting [`IPUCompiler.DISABLE_PRINT`](@ref):
 ```julia
 $(@__MODULE__).DISABLE_PRINT[] = true
 ```
@@ -300,7 +300,7 @@ IPU analogue of `Base.@show`. It comes with the same type restrictions as [`@ipu
 @ipushow x
 ```
 
-Printing can be completely disabled by setting
+Printing can be completely disabled by setting [`IPUCompiler.DISABLE_PRINT`](@ref):
 ```julia
 $(@__MODULE__).DISABLE_PRINT[] = true
 ```

src/compiler/runtime.jl

@@ -49,6 +49,9 @@ Base.rand(T::Type{Float16}) = ccall("extern _llvm_colossus_urand_f16",  llvmcall
 Base.rand(T::Type{Float32}) = ccall("extern _llvm_colossus_urand_f32",  llvmcall, Float32, ()) + T(0.5)
 Base.rand(T::Type{UInt32}) = ccall("extern _llvm_colossus_urand32",  llvmcall, UInt32, ()) + T(0.5)
 Base.rand(T::Type{UInt64}) = ccall("extern _llvm_colossus_urand64",  llvmcall, UInt64, ()) + T(0.5)
+# Note: `llvm.colossus.f{16,32}v2grand` return 2-tuples of numbers, but Julia's `Base.randn`
+# returns a single number at a time, sadly we have to discard one of the numbers to keep the
+# same semantic.
 Base.randn(T::Type{Float16}) = @inbounds ccall("extern _llvm_colossus_f16v2grand",  llvmcall, NTuple{2, VecElement{Float16}}, ())[1].value
 Base.randn(T::Type{Float32}) = @inbounds ccall("extern _llvm_colossus_f32v2grand",  llvmcall, NTuple{2, VecElement{Float32}}, ())[1].value